An error occurred fetching the project authors.
- 12 Apr, 2024 3 commits
-
-
Ian Rogers authored
Move more functionality into dsos.c generally from machine.c, renaming functions to match their new usage. The find function is made to always "get" before returning a dso. Reduce the scope of locks in vdso to match the locking paradigm. Signed-off-by:
Ian Rogers <irogers@google.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Anne Macedo <retpolanne@posteo.net> Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com> Cc: Ben Gainey <ben.gainey@arm.com> Cc: Changbin Du <changbin.du@huawei.com> Cc: Chengen Du <chengen.du@canonical.com> Cc: Colin Ian King <colin.i.king@gmail.com> Cc: Ilkka Koskinen <ilkka@os.amperecomputing.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Clark <james.clark@arm.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: K Prateek Nayak <kprateek.nayak@amd.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Leo Yan <leo.yan@linux.dev> Cc: Li Dong <lidong@vivo.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Markus Elfring <Markus.Elfring@web.de> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paran Lee <p4ranlee@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ravi Bangoria <ravi.bangoria@amd.com> Cc: Song Liu <song@kernel.org> Cc: Sun Haiyong <sunhaiyong@loongson.cn> Cc: Thomas Richter <tmricht@linux.ibm.com> Cc: Yang Jihong <yangjihong1@huawei.com> Cc: Yanteng Si <siyanteng@loongson.cn> Cc: Yicong Yang <yangyicong@hisilicon.com> Cc: zhaimingbing <zhaimingbing@cmss.chinamobile.com> Link: https://lore.kernel.org/r/20240410064214.2755936-3-irogers@google.comSigned-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Ian Rogers authored
Move functions from machine and build-id to dsos. Pass 'struct dsos' rather than internal state. Rename some functions to better represent which data structure they operate on. Signed-off-by:
-
James Clark authored
In a debug build there is validation that mmap lists are sorted when taking a lock. In machine__update_kernel_mmap() the start and end addresses are updated resulting in an unsorted list before the map is removed from the list. When the map is removed, the lock is taken which triggers the validation and the failure: $ perf test "object code reading" --- start --- perf: util/maps.c:88: check_invariants: Assertion `map__start(prev) <= map__start(map)' failed. Aborted Fix it by updating the addresses after removal, but before insertion. The bug depends on the ordering and type of debug info on the system and doesn't reproduce everywhere. Fixes: 659ad349 ("perf maps: Switch from rbtree to lazily sorted array for addresses") Reviewed-by:
James Clark <james.clark@arm.com> Acked-by:
Namhyung Kim <namhyung@kernel.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com> Cc: Ian Rogers <irogers@google.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Leo Yan <leo.yan@linux.dev> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Spoorthy S <spoorts2@in.ibm.com> Link: https://lore.kernel.org/r/20240410103458.813656-4-james.clark@arm.comSigned-off-by:
-
- 21 Mar, 2024 1 commit
-
-
Anne Macedo authored
The 'perf lock contention' program currently shows the caller of the locks as __traceiter_contention_begin+0x??. This caller can be ignored, as it is from the traceiter itself. Instead, it should show the real callers for the locks. When fiddling with the --stack-skip parameter, the actual callers for the locks start to show up. However, just ignore the __traceiter_contention_begin and the __traceiter_contention_end symbols so the actual callers will show up. Before this patch is applied: sudo perf lock con -a -b -- sleep 3 contended total wait max wait avg wait type caller 8 2.33 s 2.28 s 291.18 ms rwlock:W __traceiter_contention_begin+0x44 4 2.33 s 2.28 s 582.35 ms rwlock:W __traceiter_contention_begin+0x44 7 140.30 ms 46.77 ms 20.04 ms rwlock:W __traceiter_contention_begin+0x44 2 63.35 ms 33.76 ms 31.68 ms mutex trace_contention_begin+0x84 2 46.74 ms 46.73 ms 23.37 ms rwlock:W __traceiter_contention_begin+0x44 1 13.54 us 13.54 us 13.54 us mutex trace_contention_begin+0x84 1 3.67 us 3.67 us 3.67 us rwsem:R __traceiter_contention_begin+0x44 Before this patch is applied - using --stack-skip 5 sudo perf lock con --stack-skip 5 -a -b -- sleep 3 contended total wait max wait avg wait type caller 2 2.24 s 2.24 s 1.12 s rwlock:W do_epoll_wait+0x5a0 4 1.65 s 824.21 ms 412.08 ms rwlock:W do_exit+0x338 2 824.35 ms 824.29 ms 412.17 ms spinlock get_signal+0x108 2 824.14 ms 824.14 ms 412.07 ms rwlock:W release_task+0x68 1 25.22 ms 25.22 ms 25.22 ms mutex cgroup_kn_lock_live+0x58 1 24.71 us 24.71 us 24.71 us spinlock do_exit+0x44 1 22.04 us 22.04 us 22.04 us rwsem:R lock_mm_and_find_vma+0xb0 After this patch is applied: sudo ./perf lock con -a -b -- sleep 3 contended total wait max wait avg wait type caller 4 4.13 s 2.07 s 1.03 s rwlock:W release_task+0x68 2 2.07 s 2.07 s 1.03 s rwlock:R mm_update_next_owner+0x50 2 2.07 s 2.07 s 1.03 s rwlock:W do_exit+0x338 1 41.56 ms 41.56 ms 41.56 ms mutex cgroup_kn_lock_live+0x58 2 36.12 us 18.83 us 18.06 us rwlock:W do_exit+0x338 Signed-off-by:Anne Macedo <retpolanne@posteo.net> Acked-by:
-
- 04 Mar, 2024 4 commits
-
-
Ian Rogers authored
Move threads out of machine and into its own file. Signed-off-by:
-
Ian Rogers authored
Move thread_rb_node into the machine.c file. This hides the implementation of threads from the rest of the code allowing for it to be refactored. Locking discipline is tightened up in this change. As the lock is now encapsulated in threads, the findnew function requires holding it (as it already did in machine). Rather than do conditionals with locks based on whether the thread should be created (which could potentially be error prone with a read lock match with a write unlock), have a separate threads__find that won't create the thread and only holds the read lock. This effectively duplicates the findnew logic, with the existing findnew logic only operating under a write lock assuming creation is necessary as a previous find failed. The creation may still fail with the write lock due to another thread. The duplication is removed in a later next patch that delegates the implementation to hashtable. Signed-off-by:
-
Ian Rogers authored
Avoid exposing the threads data structure by switching to the callback machine__for_each_thread approach. machine__fprintf is only used in tests and verbose >3 output so don't turn to list and sort. Add machine__threads_nr to be refactored later. Note, all existing *_fprintf routines ignore fprintf errors. Signed-off-by:
-
Ian Rogers authored
Commit 91e467bc ("perf machine: Use hashtable for machine threads") made the iteration of thread tids unordered. The perf report --tasks output now shows child threads in an order determined by the hashing. For example, in this snippet tid 3 appears after tid 256 even though they have the same ppid 2: ``` $ perf report --tasks % pid tid ppid comm 0 0 -1 |swapper 2 2 0 | kthreadd 256 256 2 | kworker/12:1H-k 693761 693761 2 | kworker/10:1-mm 1301762 1301762 2 | kworker/1:1-mm_ 1302530 1302530 2 | kworker/u32:0-k 3 3 2 | rcu_gp ... ``` The output is easier to read if threads appear numerically increasing. To allow for this, read all threads into a list then sort with a comparator that orders by the child task's of the first common parent. The list creation and deletion are created as utilities on machine. The indentation is possible by counting the number of parents a child has. With this change the output for the same data file is now like: ``` $ perf report --tasks % pid tid ppid comm 0 0 -1 |swapper 1 1 0 | systemd 823 823 1 | systemd-journal 853 853 1 | systemd-udevd 3230 3230 1 | systemd-timesyn 3236 3236 1 | auditd 3239 3239 3236 | audisp-syslog 3321 3321 1 | accounts-daemon ... ``` Signed-off-by:
-
- 12 Feb, 2024 4 commits
-
-
Ian Rogers authored
After this change maps__nr_maps is only used by tests, existing users are migrated to maps__empty. Compute maps__empty under the read lock. Signed-off-by:
-
Ian Rogers authored
Finding a map is done under a lock, returning the map without a reference count means it can be removed without notice and causing uses after free. Grab a reference count to the map within the lock region and return this. Fix up locations that need a map__put following this. Signed-off-by:
-
Ian Rogers authored
Finding a map is done under a lock, returning the map without a reference count means it can be removed without notice and causing uses after free. Grab a reference count to the map within the lock region and return this. Fix up locations that need a map__put following this. Also fix some reference counted pointer comparisons. Signed-off-by:
-
Ian Rogers authored
Finding a map is done under a lock, returning the map without a reference count means it can be removed without notice and causing uses after free. Grab a reference count to the map within the lock region and return this. Fix up locations that need a map__put following this. Signed-off-by:
-
- 20 Dec, 2023 2 commits
-
-
Ian Rogers authored
Use to remove map_rb_node use from machine.c. Signed-off-by:
-
Ian Rogers authored
Rename maps__clone() to maps__copy_from() to be more intention revealing of its behavior. Pass the underlying maps rather than the thread. Signed-off-by:
-
- 19 Dec, 2023 1 commit
-
-
Ian Rogers authored
Switch machine__map_x86_64_entry_trampolines and machine__for_each_kernel_map from loop macro maps__for_each_entry to maps__for_each_map function that takes a callback. The function holds the maps lock, which should be held during iteration. Signed-off-by:
-
- 06 Dec, 2023 1 commit
-
-
Ian Rogers authored
When mapping an IP it is either an identity mapping or a DSO relative mapping, so a single bit is required in the struct to identify this. The current code uses function pointers, adding 2 pointers per map and also pushing the size of a map beyond 1 cache line. Switch to using a byte to identify the mapping type (as well as priv and erange_warned), to avoid any masking. Change struct maps's layout to avoid holes. Before: ``` struct map { u64 start; /* 0 8 */ u64 end; /* 8 8 */ _Bool erange_warned:1; /* 16: 0 1 */ _Bool priv:1; /* 16: 1 1 */ /* XXX 6 bits hole, try to pack */ /* XXX 3 bytes hole, try to pack */ u32 prot; /* 20 4 */ u64 pgoff; /* 24 8 */ u64 reloc; /* 32 8 */ u64 (*map_ip)(const struct map *, u64); /* 40 8 */ u64 (*unmap_ip)(const struct map *, u64); /* 48 8 */ struct dso * dso; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ refcount_t refcnt; /* 64 4 */ u32 flags; /* 68 4 */ /* size: 72, cachelines: 2, members: 12 */ /* sum members: 68, holes: 1, sum holes: 3 */ /* sum bitfield members: 2 bits, bit holes: 1, sum bit holes: 6 bits */ /* last cacheline: 8 bytes */ }; ``` After: ``` struct map { u64 start; /* 0 8 */ u64 end; /* 8 8 */ u64 pgoff; /* 16 8 */ u64 reloc; /* 24 8 */ struct dso * dso; /* 32 8 */ refcount_t refcnt; /* 40 4 */ u32 prot; /* 44 4 */ u32 flags; /* 48 4 */ enum mapping_type mapping_type:8; /* 52: 0 4 */ /* Bitfield combined with next fields */ _Bool erange_warned; /* 53 1 */ _Bool priv; /* 54 1 */ /* size: 56, cachelines: 1, members: 11 */ /* padding: 1 */ /* last cacheline: 56 bytes */ }; ``` Signed-off-by:
-
- 09 Nov, 2023 1 commit
-
-
Ian Rogers authored
'struct thread' values hold onto references to mmaps, DSOs, etc. When a thread exits it is necessary to clean all of this memory up by removing the thread from the machine's threads. Some tools require this doesn't happen, such as auxtrace events, 'perf report' if offcpu events exist or if a task list is being generated, so add a 'struct symbol_conf' member to make the behavior optional. When an exited thread is left in the machine's threads, mark it as exited. This change relates to commit 40826c45 ("perf thread: Remove notion of dead threads") . Dead threads were removed as they had a reference count of 0 and were difficult to reason about with the reference count checker. Here a thread is removed from threads when it exits, unless via symbol_conf the exited thread isn't remove and is marked as exited. Reference counting behaves as it normally does. Reviewed-by:
Adrian Hunter <adrian.hunter@intel.com> Signed-off-by:
-
- 25 Oct, 2023 4 commits
-
-
Ian Rogers authored
Fix leak where mem_info__put wouldn't release the maps/map as used by perf mem. Add exit functions and use elsewhere that the maps and map are released. Signed-off-by:
-
Ian Rogers authored
Commit 40826c45 ("perf thread: Remove notion of dead threads") removed dead threads but the list head wasn't removed. Remove it here. Signed-off-by:
-
Ian Rogers authored
Comparing pointers with reference count checking is tricky to avoid a SEGV. Add a convenience macro to simplify and use. Signed-off-by:
-
Ian Rogers authored
Running perf top with address sanitizer and "--call-graph=lbr" fails due to reading sample 0 when no samples exist. Add a guard to prevent this. Fixes: e2b23483 ("perf machine: Factor out lbr_callchain_add_lbr_ip()") Signed-off-by:
-
- 11 Sep, 2023 1 commit
-
-
Jiapeng Chong authored
Fix the following coccicheck warnings: ./tools/perf/util/machine.c:2000:9-10: WARNING: return of 0/1 in function 'symbol__match_regex' with return type bool. Committer notes: Found this in the pile, it was already returning bool, but this patch simplifies it further, from 3 lines to just 1. Reported-by:
Abaci Robot <abaci@linux.alibaba.com> Signed-off-by:
Jiapeng Chong <jiapeng.chong@linux.alibaba.com> Suggested-by:
David Laight <David.Laight@ACULAB.COM> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Andrii Nakryiko <andrii@kernel.org> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: John Fastabend <john.fastabend@gmail.com> Cc: KP Singh <kpsingh@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Martin KaFai Lau <kafai@fb.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Song Liu <songliubraving@fb.com> Cc: Yonghong Song <yhs@fb.com> Cc: bpf@vger.kernel.org Cc: netdev@vger.kernel.org Link: https://lore.kernel.org/r/1614247483-102665-1-git-send-email-jiapeng.chong@linux.alibaba.comSigned-off-by:
-
- 08 Aug, 2023 1 commit
-
-
Arnaldo Carvalho de Melo authored
This reverts commit 46d21ec0. The tests were made with a specific workload, further tests on a recently updated fedora 38 system with a system wide perf.data file shows 'perf report' taking excessive time resolving inlines in vmlinux, so lets revert this until a full investigation and improvement on the addr2line support code is made. Reported-by:
Jesper Dangaard Brouer <hawk@kernel.org> Acked-by:
Artem Savkov <asavkov@redhat.com> Tested-by:
-
- 28 Jul, 2023 1 commit
-
-
Namhyung Kim authored
When 'perf record -d' is used, it needs data mmaps to symbolize global data. But it missed to collect kernel data maps so it cannot symbolize them. Instead of having a separate map, just increase the kernel map size to include the data section. Probably we can have a separate kernel map for data, but the current code assumes a single kernel map. So it'd require more changes in other places and looks error-prone. I decided not to go that way for now. Also it seems the kernel module size already includes the data section. For example, my system has the following. $ grep -e _stext -e _etext -e _edata /proc/kallsyms ffffffff99800000 T _stext ffffffff9a601ac8 T _etext ffffffff9b446a00 D _edata Size of the text section is (0x9a601ac8 - 0x99800000 = 0xe01ac8) and size including data section is (0x9b446a00 - 0x99800000 = 0x1c46a00). Before: $ perf record -d true $ perf report -D | grep MMAP | head -1 0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0xe01ac8) @ 0xffffffff99800000]: x [kernel.kallsyms]_text ^^^^^^^^ here After: $ perf report -D | grep MMAP | head -1 0 0 0x460 [0x60]: PERF_RECORD_MMAP -1/0: [0xffffffff99800000(0x1c46a00) @ 0xffffffff99800000]: x [kernel.kallsyms]_text ^^^^^^^^^ Instead of just replacing it to _edata, try _edata first and then fall back to _etext just in case. Signed-off-by:
-
- 12 Jun, 2023 11 commits
-
-
Ian Rogers authored
Pthread keys are more portable than __thread and allow the association of a destructor with the key. Use the destructor to clean up TLS callchain cursors to aid understanding memory leaks. Committer notes: Had to fixup a series of unconverted places and also check for the return of get_tls_callchain_cursor() as it may fail and return NULL. In that unlikely case we now either print something to a file, if the caller was expecting to print a callchain, or return an error code to state that resolving the callchain isn't possible. In some cases this was made easier because thread__resolve_callchain() already can fail for other reasons, so this new one (cursor == NULL) can be added and the callers don't have to explicitely check for this new condition. Signed-off-by:
-
Ian Rogers authored
Fix missed reference count gets and puts as detected with leak sanitizer and reference count checking. Signed-off-by:
-
Ian Rogers authored
machine__addnew_module_map requires a put on its result. Add this and narrow the scope of map to make the correctness more obvious. This leak was caught with leak sanitizer and the reference count checker. Signed-off-by:
-
Ian Rogers authored
The kernel dso may be found by searching dsos or allocating if not found. The allocation returns with a reference count of 2, once for the dsos list and once for the returned value. The list search has a reference count of 1, once for the dsos list. To make the reference counts consistent, increase the dsos list search reference count to 2 with a dso__get, and do a put when the scope ends for either the allocated or found dso. This issue was found with leak sanitizer and reference count checking. Signed-off-by:
-
Ian Rogers authored
The code required threads to be deleted before machine__exit was called or the threads would be leaked. This was error prone so move the delete_threads into machine__exit. Signed-off-by:
-
Ian Rogers authored
Modify struct declaration and accessor functions for the reference count checkers additional layer of indirection. Make sure pid_cmp in builtin-sched.c uses the underlying/original struct in pointer arithmetic, and not the temporary get/put indirection. Signed-off-by:
-
Ian Rogers authored
struct addr_location holds references to multiple reference counted objects. Add init/exit functions to make maintenance of those more consistent with the rest of the code and to try to avoid leaks. Modification of thread reference counts isn't included in this change. Committer notes: I needed to initialize result to sample->ip to make sure is set to something, fixing a compile time error, mostly keeping the previous logic as build_alloc_func_list() already does debugging/error prints about what went wrong if it takes the 'goto out'. Signed-off-by:
-
Ian Rogers authored
Address/leak sanitizer with reference count checking can identify the location of leaks, so use put rather than delete to avoid free-ing memory when the reference count is >1. Add maps__zput to ensure the variable is cleared. Signed-off-by:
-
Ian Rogers authored
Using accessors will make it easier to add reference count checking in later patches. Committer notes: thread->nsinfo wasn't wrapped as it is used together with nsinfo__zput(), where does a trick to set the field with a refcount being dropped to NULL, and that doesn't work well with using thread__nsinfo(thread), that loses the &thread->nsinfo pointer. When refcount checking is added to 'struct thread', later in this series, nsinfo__zput(RC_CHK_ACCESS(thread)->nsinfo) will be used to check the thread pointer. Signed-off-by:
-
Ian Rogers authored
Separate the rbtree out of thread and into a new struct thread_rb_node. The refcnt is in thread and the rbtree is responsible for a single count. Signed-off-by:
-
Ian Rogers authored
The dead thread list is best effort. Threads live on it until the reference count hits zero and they are removed. With correct reference counting this should never happen. It is, however, part of the 'perf sched' output that is now removed. If this is an issue we should implement tracking of dead threads in a robust not best-effort way. Signed-off-by:
-
- 19 Apr, 2023 3 commits
-
-
Ian Rogers authored
There's no strict get/put policy with map that leads to leaks or use after free. Reference count checking identifies correct pairing of gets and puts. Committer notes: Extracted from a larger patch removing bits that were covered by the use of pre-existing map__ accessors (e.g. maps__nr_maps()) and new ones added (map__refcnt() and the maps__set_ ones) to reduce RC_CHK_ACCESS(maps)-> source code pollution. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
perf map: Add set_ methods for map->{start,end,pgoff,pgoff,reloc,erange_warned,dso,map_ip,unmap_ip,priv} To have a way to intercept usage of the reference counted struct map. Signed-off-by: -
Arnaldo Carvalho de Melo authored
To use the existing accessor and be consistent. Signef-off-by:
-
- 07 Apr, 2023 2 commits
-
-
Ian Rogers authored
When a pointer to a map exists do a get, when that pointer is overwritten or freed, put the map. This avoids issues with gets and puts being inconsistently used causing, use after puts, etc. For example, the map in struct addr_location is changed to hold a reference count. Reference count checking and address sanitizer were used to identify issues. Signed-off-by:
-
Ian Rogers authored
Later changes will add reference count checking for 'struct map'. Add accessors so that the reference count check is only necessary in one place. Signed-off-by:
-
