Em Fri, Mar 27, 2015 at 11:08:00AM +0900, Namhyung Kim escreveu: > Hello, > > Currently perf kmem command only analyzes SLAB memory allocation. And > I'd like to introduce page allocation analysis also. Users can use > --slab and/or --page option to select it. If none of these options > are used, it does slab allocation analysis for backward compatibility. > > * changes in v5) > - print migration type and gfp flags in more compact form (Arnaldo) > - add kmem.default config option Ok, I'll try this after lunch and if nobody else has any more issues with it, merge it, OK? Ingo? - Arnaldo > * changes in v4) > - use pfn instead of struct page * in tracepoints (Joonsoo, Ingo) > - print gfp flags in human readable string (Joonsoo, Minchan) > > * changes in v3) > - add live page statistics > > * changes in v2) > - Use thousand grouping for big numbers - i.e. 12345 -> 12,345 (Ingo) > - Improve output stat readability (Ingo) > - Remove alloc size column as it can be calculated from hits and order > > Patch 1 is to convert tracepoint to save pfn instead of struct page *. > Patch 2 implements basic support for page allocation analysis, patch 3 > deals with the callsite and patch 4 implements sorting. Patch 5 > introduces live page analysis which is to focus on currently allocated > pages only. Patch 6 prints gfp flags in human readable string and > patch 7 adds kmem.default config option to be able to select page > analysis by default. > > In this patchset, I used two kmem events: kmem:mm_page_alloc and > kmem_page_free for analysis as they can track almost all of memory > allocation/free path AFAIK. However, unlike slab tracepoint events, > those page allocation events don't provide callsite info directly. So > I recorded callchains and extracted callsites like below: > > Normal page allocation callchains look like this: > > 360a7e __alloc_pages_nodemask > 3a711c alloc_pages_current > 357bc7 __page_cache_alloc <-- callsite > 357cf6 pagecache_get_page > 48b0a prepare_pages > 494d3 __btrfs_buffered_write > 49cdf btrfs_file_write_iter > 3ceb6e new_sync_write > 3cf447 vfs_write > 3cff99 sys_write > 7556e9 system_call > f880 __write_nocancel > 33eb9 cmd_record > 4b38e cmd_kmem > 7aa23 run_builtin > 27a9a main > 20800 __libc_start_main > > But first two are internal page allocation functions so it should be > skipped. To determine such allocation functions, I used following regex: > > ^_?_?(alloc|get_free|get_zeroed)_pages? > > This gave me a following list of functions (you can see this with -v): > > alloc func: __get_free_pages > alloc func: get_zeroed_page > alloc func: alloc_pages_exact > alloc func: __alloc_pages_direct_compact > alloc func: __alloc_pages_nodemask > alloc func: alloc_page_interleave > alloc func: alloc_pages_current > alloc func: alloc_pages_vma > alloc func: alloc_page_buffers > alloc func: alloc_pages_exact_nid > > After skipping those function, it got '__page_cache_alloc'. > > Other information such as allocation order, migration type and gfp > flags are provided by tracepoint events. > > Basically the output will be sorted by total allocation bytes, but you > can change it by using -s/--sort option. The following sort keys are > added to support page analysis: page, order, mtype, gfp. Existing > 'callsite', 'bytes' and 'hit' sort keys also can be used. > > An example follows: > > # perf kmem record --page sleep 5 > [ perf record: Woken up 2 times to write data ] > [ perf record: Captured and wrote 1.065 MB perf.data (2949 samples) ] > > # perf kmem stat --page > # GFP flags > # --------- > # 00000010: NI: GFP_NOIO > # 000000d0: K: GFP_KERNEL > # 00000200: NWR: GFP_NOWARN > # 000052d0: K|NWR|NR|C: GFP_KERNEL|GFP_NOWARN|GFP_NORETRY|GFP_COMP > # 000084d0: K|R|Z: GFP_KERNEL|GFP_REPEAT|GFP_ZERO > # 000200d0: U: GFP_USER > # 000200d2: HU: GFP_HIGHUSER > # 000200da: HUM: GFP_HIGHUSER_MOVABLE > # 000280da: HUM|Z: GFP_HIGHUSER_MOVABLE|GFP_ZERO > # 002084d0: K|R|Z|NT: GFP_KERNEL|GFP_REPEAT|GFP_ZERO|GFP_NOTRACK > # 0102005a: NF|HW|M: GFP_NOFS|GFP_HARDWALL|GFP_MOVABLE > > --------------------------------------------------------------------------------------------------------- > Total alloc (KB) | Hits | Order | Mig.type | GFP flags | Callsite > --------------------------------------------------------------------------------------------------------- > 3,876 | 969 | 0 | MOVABLE | HUM | shmem_alloc_page > 972 | 243 | 0 | UNMOVABL | K | __pollwait > 624 | 156 | 0 | MOVABLE | NF|HW|M | __page_cache_alloc > 304 | 76 | 0 | UNMOVABL | U | dma_generic_alloc_coherent > 108 | 27 | 0 | MOVABLE | HUM|Z | handle_mm_fault > 56 | 14 | 0 | UNMOVABL | K|R|Z|NT | pte_alloc_one > 24 | 6 | 0 | MOVABLE | HUM | do_wp_page > 24 | 3 | 1 | UNMOVABL | K|NWR|NR|C | alloc_skb_with_frags > 16 | 4 | 0 | UNMOVABL | NWR | __tlb_remove_page > 16 | 4 | 0 | MOVABLE | HUM | do_cow_fault > ... | ... | ... | ... | ... | ... > --------------------------------------------------------------------------------------------------------- > > SUMMARY (page allocator) > ======================== > Total allocation requests : 1,518 [ 6,096 KB ] > Total free requests : 1,431 [ 5,748 KB ] > > Total alloc+freed requests : 1,330 [ 5,344 KB ] > Total alloc-only requests : 188 [ 752 KB ] > Total free-only requests : 101 [ 404 KB ] > > Total allocation failures : 0 [ 0 KB ] > > Order Unmovable Reclaimable Movable Reserved CMA/Isolated > ----- ------------ ------------ ------------ ------------ ------------ > 0 351 . 1,163 . . > 1 3 . . . . > 2 1 . . . . > 3 . . . . . > 4 . . . . . > 5 . . . . . > 6 . . . . . > 7 . . . . . > 8 . . . . . > 9 . . . . . > 10 . . . . . > > > I have some idea how to improve it. But I'd also like to hear other > idea, suggestion, feedback and so on. > > This is available at perf/kmem-page-v5 branch on my tree: > > git://git.kernel.org/pub/scm/linux/kernel/git/namhyung/linux-perf.git > > Thanks, > Namhyung > > > Namhyung Kim (7): > tracing, mm: Record pfn instead of pointer to struct page > perf kmem: Analyze page allocator events also > perf kmem: Implement stat --page --caller > perf kmem: Support sort keys on page analysis > perf kmem: Add --live option for current allocation stat > perf kmem: Print gfp flags in human readable string > perf kmem: Add kmem.default config option > > include/trace/events/filemap.h | 8 +- > include/trace/events/kmem.h | 42 +- > include/trace/events/vmscan.h | 8 +- > tools/perf/Documentation/perf-kmem.txt | 19 +- > tools/perf/builtin-kmem.c | 1298 ++++++++++++++++++++++++++++++-- > 5 files changed, 1288 insertions(+), 87 deletions(-) > > -- > 2.3.4 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>