Commit 6d23d3ac authored by Oran Agra's avatar Oran Agra
Browse files

Squashed 'deps/jemalloc/' changes from ea6b3e973..54eaed1d8

54eaed1d8 Merge branch 'dev'
304c91982 Update ChangeLog for 5.3.0.
8cb814629 Make the default option of zero realloc match the system allocator.
66c889500 Make test/unit/background_thread_enable more conservative.
a7d73dd4c Update TUNING.md to include the new tcache_max option.
254b01191 Small doc tweak of opt.trust_madvise.
f5e840bbf Minor typo fix in doc.
ceca07d2c Correct the name of stats.mutexes.prof_thds_data in doc.
391bad4b9 Avoid abort() in test/integration/cpp/infallible_new_true.
9a242f16d fix some typos
0e29ad4ef Rename zero_realloc option "strict" to "alloc".
5841b6dbe Update FreeBSD image to 12.3 for cirrus ci.
ed5fc14b2 Use volatile to workaround buffer overflow false positives.
25517b852 Reoreder TravisCI jobs to optimize CI time
8a49b62e7 Enable TravisCI for Windows
fdb6c1016 Add FreeBSD to TravisCI
a93931537 Do not disable SEC by default for 64k pages platforms
eaaa368ba Add comments and use meaningful vars in sz_psz2ind.
5bf03f8ce Implement PAGE_FLOOR macro
52631c90f Fix size class calculation for sec
7ae0f15c5 Add a default page size when cross-compile for Apple M1.
eb65d1b07 Fix FreeBSD system jemalloc TSD cleanup
78b58379c Fix possible "nmalloc >= ndalloc" assertion.
ca709c313 Fix failed assertion due to racy memory access
063d134ae Properly detect background thread support on Darwin.
a4e81221c Document 'make uninstall'
20f9802e4 Avoid overflow warnings in test/unit/safety_check.
8c59c44ff Add a dependency checking step at the end of malloc_conf_init.
efc539c04 Initialize prof_leak during prof init.
002f0e939 Disable TravisCI jobs generation for Windows
01a293fc0 Add Windows to TravisCI
b798fabdf Add prof_leak_error option
eafd2ac39 Forbid spaces in prefix and exec_prefix
36a09ba2c Forbid spaces in install suffix
640c3c72e Add support for 'make uninstall'
f15d8f3b4 Echo installed files via verbose 'install' command
eb196815d Avoid calculating size of size class twice & delete sc_data_global.
011449f17 Fix doc build with install-suffix.
8b49eb132 Fix the HELP_STRING of --enable-doc.
ddb170b1d Simplify arena_migrate() to take arena_t* instead of indices.
648b3b9f7 Lower the num_threads in the stress test of test/unit/prof_recent
d66162e03 Fix the extent state checking on the merge error path.
c9946fa7e FreeBSD also needs the OS-X "don't declare system functions as nothrow" fix since it also has jemalloc in the base system
89fe8ee6b Use the isb instruction instead of yield for spin locks on arm
6230cc88b Add background thread sleep retry in test/unit/hpa_background_thread
61978bbe6 Purge all if the last thread migrated away from an arena.
c91e62dd3 #include <features.h> as requested
18510020e Fix symbol conflict with musl libc
f509703af Fix two conversion warnings in tcache.
067c2da07 Fix unnecessary returns in san_(un)guard_pages_two_sided.
d660683d3 Fix test config of lg_san_uaf_align.
eabe88916 Rename full_position to low_bound in cache_bin.h.
dfdd7562f Rename san_enabled() to san_guard_enabled().
01d61a3c6 Fix a conversion warning.
8b34a788b Fix an used-uninitialized warning (false positive).
e491cef9a Add stats for stashed bytes in tcache.
b75822bc6 Implement use-after-free detection using junk and stash.
06aac61c4 Split the core logic of tcache flush into a separate function.
d038160f3 Fix shadowed variable usage.
bd70d8fc0 Add the profiling settings for tests explicit.
e491df1d2 Fix warnings when using autoheader.
60b9637cc Only invoke malloc_cpu_count_is_deterministic() when necessary.
837b37c4c Fix the time-since computation in HPA.
310af725b Add nstime_ns_since which obtains the duration since the input time.
cafe9a315 Disable percpu arena in case of non deterministic CPU count
bb5052ce9 Fix base_ehooks_get_for_metadata
9015e129b Update visual studio projects
d90655390 San: Create a function for committing and zeroing
800ce49c1 San: Bump alloc frequently reused guarded allocations
f56f5b993 Pass 'frequent_reuse' hint to PAI
2c70e8d35 Rename 'arena_decay' to 'arena_util'
0f6da1257 San: Implement bump alloc
34b00f896 San: Avoid running san tests with prof enabled
62f9c54d2 San: Rename 'guard' to 'san'
d9bbf539f CI: Refactor gen_travis.py
7dcf77809 Mark slab as true on sized dealloc fast path.
af6ee27c0 Enforce abort_conf:true when malloc_conf is not fully recognized.
113e8e68e freebsd 14 build fix proposal.
3b3257a70 Correct opt.prof_leak documentation
cdabe908d Track the initialized state of nstime_t on debug build.
400c59895 Fix uninitialized nstime reading / updating on the stack in hpa.
8b81d3f21 Fix the initialization of last_event in thread event init.
6bdb4f5ab Check prof_active in addtion to opt_prof during batch_alloc().
37342a4d3 Add ctl interface for experimental_infallible_new.
6cb585b13 San: Unguard guarded slabs during arena destruction
b6a7a535b Optimize away a branch on the free fastpath.
4d56aaeca Optimize away the tsd_fast() check on free fastpath.
26f5257b8 Remove declaration of an undefined function
215961541 Add new architecture loongarch.
8daac7958 Redefine functions with test hooks only for tests
c9ebff0fd Initialize deferred_work_generated
912324a1a Add debug check outside of the loop in hpa_alloc_batch.
cf9724531 Darwin malloc_size override support proposal.
ab0f1604b Delay the atexit call to prof_log_start().
11b6db744 CPU affinity on BSD platforms support.
83f329402 Small refactors around 7bb05e0.
3c4b717ff Remove unused header base_structs.h.
deb8e62a8 Implement guard pages.
7bb05e04b add experimental.arenas_create_ext mallctl
a9031a097 Allow setting a dump hook
f7d46b811 Allow setting custom backtrace hook
523cfa55c Guard prof related mallctl with opt_prof.
6e848a005 Remove opt_background_thread_hpa_interval_max_ms
8229cc77c Wake up background threads on demand
97da57c13 HPA: Add min_purge_interval_ms option
b8b8027f1 Allow PAI to calculate time until deferred work
26140dd24 Reject --enable-prof-libunwind without --enable-prof
e5062e9fb Makefile.in: make sure doc generated before install
8b24cb8fd Don't assume initialized arena in the default alloc hook.
c01a885e9 HPA: Correctly calculate retained pages
2c625d5cd Fix warnings when compiled with clang
9d02bdc88 Port gen_run_tests.py to python3
5884a076f Rename prof.dump_prefix to prof.prefix
6a0160071 Add Cirrus CI testing matrix
f58064b93 Verify that HPA is used before calling its functions
27f71242b Mutex: Tweak internal spin count.
6f41ba55e Mutex: Make spin count configurable.
dae24589b PH: Insert-below-min fast-path.
40d53e007 ph: Add aux-list counting and pre-merging.
dcb7b83fa Eset: Cache summary information for heap edatas.
252e0942d Eset: Pull per-pszind data into structs.
dc0a4b8b2 Edata: Pull out comparison fields into a summary.
0170dd198 Edata: Fix a couple typos.
08a4cc096 Pairing heap: inline functions instead of macros.
92a1e38f5 edata_cache: Allow unbounded fast caching.
d93eef2f4 HPA: Introduce a redesigned hpa_central_t.
e09eac1d4 Remove hpa_central.
c88fe355e Add unit tests for decay
aaea4fd1e Add more documentation to decay.c
4b633b9a8 Clean up background thread sleep computation
6630c5989 HPA: Hugification hysteresis.
113938b6f HPA: Pull out a hooks type.
1d4a7666d HPA: Do deferred operations on background threads.
583284f2d Add HPA deferral functionality.
ace329d11 HPA batch dalloc: Just do one deferred work check.
47d8a7e6b psset: Purge empty slabs first.
41fd56605 HPA: Purge across retained extents.
347523517 PAI: Fix a typo.
9c42ed2d1 Travis: Don't test "clang" on OS X.
d202218e8 HPA: Fix typos with big performance implications.
de033f56c mpsc_queue: Add module.
4452a4812 Add opt.experimental_infallible_new.
0689448b1 Travis: Unbreak the builds.
4fb93a18e extent_can_acquire_neighbor typo fix
2381efab5 ARC: add Minimum allocation alignment
2c0f4c2ac Fix typo in configure.ac: experimetal -> experimental
36c6bfb96 SEC: Allow arbitrarily many shards, cached sizes.
11beab38b Added --debug-syms-by-id option
08089589f Fix an interaction between the oversize_threshold test and bgthds.
541793821 Red-black tree: add summarize/filter.
b2c08ef2e RB unit tests: don't test reentrantly.
aea91b8c3 Clean up some minor data structure inconsistencies
1f688490e Stats: Fix a printing bug when hpa_dirty_mult = -1
4f7cb3a41 Sized deallocation: fix a typo.
12cd13cd4 Fix thread.name/prof_sys_thread_name interaction
304cdbb13 Fix a prof_recent/prof_sys_thread_name interaction
9b523c6c1 Refactor the locking in extent_recycle().
ce68f326b Avoid the release & re-acquire of the ecache locks around the merge hook.
7dc77527b Delete the mutex_pool module.
03d95cba8 Remove the unnecessary arena_ind_set in base_alloc_edata().
3093d9455 Move the edata mergeability related functions to extent.h.
7c964b035 Add rtree_write_range(): writing the same content to multiple leaf elements.
add636596 Stop checking head state in the merge hook.
49b7d7f0a Passing down the original edata on the expand path.
178493968 Use rtree tracked states to protect edata outside of ecache locks.
9ea235f8f Add witness_assert_positive_depth_to_rank().
4d8c22f9a Store edata->state in rtree leaf and make edata_t 128B aligned.
70d1541c5 Track extent is_head state in rtree leaf.
862219e46 Add quiescence sync before deleting base during arena_destroy.
a137a6825 Remove redundant declaration, pac_retain_grow_limit_get_set was declared twice in pac.h
2ae1ef7db Fix doc large size 54 KiB error
61afb6a40 Fix locking on arena_i_destroy_ctl().
9193ea224 Cirrus: fix build.
391307714 Mark head state during dss alloc.
11127240c Remove redundant enable-debug definition in configure.
22be724af Set is_head in extent_alloc_wrapper w/ retain.
73ca4b8ef HPA: Use dirtiest-first purging.
0f6c420f8 HPA: Make purging/hugifying more principled.
6bddb92ad psset: Rename "bitmap" to "pageslab_bitmap".
154aa5fcc Use the flat bitmap for eset and psset bitmaps.
271a676dc hpdata: early bailout for longest free range.
d21d5b46b Edata: Move sn into its own field.
fb327368d SEC: Expand option configurability.
ce9386370 HPA: Implement batch allocation.
cdae6706a SEC: Use batch fills.
480f3b11c Add a batch allocation interface to the PAI.
bf448d7a5 SEC: Reduce lock hold times.
1944ebbe7 HPA: Implement batch deallocation.
f47b4c2cd PAI/SEC: Add a dalloc_batch function.
4b8870c7d SEC: Fix a comment typo.
cde7097ec Update INSTALL.md to mention 'autoconf'
a11be5033 Implement opt.cache_oblivious.
8c5e5f50a Fix stats for "tcache_max" (was "lg_tcache_max")
041145c27 Report the correct and wrong sizes on sized dealloc bug detection.
f3b2668b3 Report the offending pointer on sized dealloc bug detection.
edbfe6912 Inline malloc fastpath into operator new.
79f81a373 HPA: Make dirty_mult configurable.
32dd15379 HPA: Make dehugification threshold configurable.
4790db15e HPA: make the hugification threshold configurable.
b3df80bc7 Pull HPA options into a containing struct.
bdb7307ff fxp: Add FXP_INIT_PERCENT
caef4c286 FXP: add fxp_mul_frac.
56e85c0e4 HPA: Use a whole-shard purging heuristic.
dc886e560 hpdata: Return the number of pages to be purged.
9fd9c876b psset: keep aggregate stats.
da63f23e6 HPA: Track pending purges/hugifies in the psset.
0ea3d6307 CTL, Stats: report HPA empty slab stats.
bf64557ed Move empty slab tracking to the psset.
99fc0717e psset: Reconceptualize insertion/removal.
061cabb71 HPA stats: report retained instead of inactive.
d3e5ea03c HPA: Track dirty stats.
68a1666e9 hpdata: Rename "dirty" to "touched".
be0d7a53f HPA: Don't track inactive pages.
55e0f60ca psset stats: Simplify handling.
94cd9444c HPA: Some minor reformattings.
b25ee5d88 HPA: Add purge stats.
746ea3de6 HPA stats: Allow some derived stats.
30b9e8162 HPA: Generalize purging.
70692cfb1 hpdata: Add state changing helpers.
9b75808be flat bitmap: Add a bitwise and/or/not.
2ae966222 hpdata: track per-page dirty state.
ff4086aa6 hpdata: count active pages instead of free ones.
3624dd42f hpdata: Add a comment for hpdata_consistent.
20140629b Bin: Move stats closer to the mutex.
c259323ab Use ticker_geom_t for arena tcache decay.
8edfc5b17 Add ticker_geom_t.
396732981 Arena: share bin offsets in a global.
2fcbd1811 Cache bin: Don't reverse flush order.
4c46e1136 Cache an arena's index in the arena.
229994a20 Tcache flush: keep common path state in registers.
31a629c3d Tcache flush: prefetch edata contents.
9f9247a62 Tcache fluhing: increase cache miss parallelism.
181ba7fd4 Tcache flush: Add an emap "batch lookup" path.
c007c537f Tcache flush: Unify edata lookup path.
35a855260 Mac OS: Tag mapped pages.
f6699803e Fix duration in prof log
a943172b7 Add runtime detection for MADV_DONTNEED zeroes pages (mostly for qemu)
2e3104ba0 Update config.{sub,guess} to support support-aarch64-apple-darwin as a target
a011c4c22 cache_bin: Separate out local and remote accesses.
14d689c0f Add prof stats mutex stats
9f71b5779 Output prof stats in stats print
1f1a0231e Split macros for initializing stats headers
4352cbc21 Add alignment tests for prof stats
54f3351f1 Add mallctl for prof stats fetching
40fa4d29d Track per size class internal fragmentation
afa489c3c Record request size in prof info
f9bb8dede Un-force-inline do_rallocx.
a9fa2defd Add JEMALLOC_COLD, and mark some functions cold.
5d8e70ab2 prof_recent: cassert(config_prof) more often.
83cad746a prof_log: cassert(config_prof) in public functions
526180b76 Extent.c: Avoid an rtree NULL-check.
b35ac00d5 Do not bump to large size for page aligned request
8a56d6b63 Add last-N mutex stats
22d62d8cb Handle ending gap properly for HPA stats
6c5a3a24d Omit bin stats rows with no data
ea013d8fa Enforce realloc sizing stability
74bd63b20 Optimize stats print using partial name-to-mib
4557c0a67 Enable ctl on partial mib and partial name
006dd0414 Add partial name-to-mib functionality
f2e1a5be7 Do not fail on partial ctl path for ctl_nametomib()
6ab181d2b Extract node lookup given mib input
3a627b967 No need to record all nodes in ctl_lookup()
91e006c4c Enable ctl_lookup() to start from arbitrary node
063a767ff Define JEMALLOC_HAS_ALLOCA_H for QNX
4e3fe218e Use posix_madvise to purge pages when available
26c1dc5a3 Support AutoConf for posix_madvise and POSIX_MADV_DONTNEED
96a59c3bb Fix recursive malloc during bootstrap on QNX
986cbe488 Disable JEMALLOC_TLS for QNX
1e3b8636f HPA: Remove unused malloc_conf options.
e82771807 Cache mallctl mib for batch allocation stress test
0dfdd31e0 Add tiny batch size to batch allocation stress test
9522ae41d Move n_search outside of assert as reported by static analyzer
a559caf74 hpdata: Strengthen assertions.
f51948d9e psset unit test: fix a bug.
54c94c167 flat bitmap: add scount / ucount functions.
e6c057ad3 fb: implement assign in terms of a visitor.
734e72ce8 bit_util: Guarantee popcount's presence.
d9f7e6c66 hpdata: Add a test.
3ed0b4e8a HPA: Add an nevictions counter.
fffcefed3 malloc_conf: Clarify HPA options.
f7cf23aa4 psset: Relegate alloc/dalloc to test code.
f9299ca57 HPA: Use psset fit/insert/remove.
0971e1e4e hpdata: Use addr/size instead of begin/npages.
5228d869e psset: Use fit/insert/remove as basis functions.
089f8fa44 Move hpdata bitmap logic out of the psset.
ca30b5db2 Introduce hpdata_t.
4a15008cf HPA unit test: skip if unsupported.
43af63fff HPA: Manage whole hugepages at a time.
63677dde6 Pages: Statically detect if pages_huge may succeed
c1b2a7793 psset: Move in stats.
d0a991d47 psset: Add insert/remove functions.
d438296b1 narenas_ratio: Accept fractional values.
ecd39418a Add fxp: A fixed-point math library.
99c2d6c23 Backport jeprof --collapse for flamegraph generation
520b75fa2 utrace support with label based signature.
92e189be8 Add some comments to the batch allocation logic flow
d96e4525a Route batch allocation of small batch size to tcache
ac480136d Split out locality checking in batch allocation tests
be5e49f4f Add a batch mode for cache_bin_alloc()
4a65f3493 Fix a cache bin test
566c4a859 Slight changes to cache bin internal functions
9545c2cd3 Add sample interval to prof last-N dump
cf2549a14 Add a per-arena oversize_threshold.
4ca3d91e9 Rename geom_grow -> exp_grow.
b4c37a6e8 Rename edata_tree_t -> edata_avail_t.
95f0a77fd Detect pthread_getname_np explicitly.
b3c5690b7 Update config.{guess,sub} to 2020-11-07@77632d9
589638182 Use the edata_cache_small_t in the HPA.
03a604711 Edata cache small: rewrite.
c9757d9e3 HPA: Don't disable shards that were never started.
1b3ee7566 Add experimental.thread.activity_callback.
27ef02ca9 Android build fix proposal.
d2d941017 MADV_DO[NOT]DUMP support equivalence on FreeBSD.
180b84315 Appveyor: fix 404 errors.
ef6d51ed4 DragonFlyBSD build support.
bf72188f8 Allow opt.tcache_max to accept small size classes.
ea32060f9 SEC: Implement thread affinity.
d16849c91 psset: Do first-fit based on slab age.
634ec6f50 Edata: add an "age" field.
6599651ae PA: Use an SEC in fron of the HPA shard.
ea51e97bb Add SEC module: a small extent cache.
1964b0839 HPA: Add stats for the hpa_shard.
534504d4a HPA: add size-exclusion functionality.
484f04733 HPA: Add central mutex contention stats.
bf025d2ec HPA: Make slab sizes and maxes configurable.
1c7da3331 HPA: Tie components into a PAI implementation.
c8209150f Switch from opt.lg_tcache_max to opt.tcache_max
5ba861715 Add thread name in prof last-N records
4ef5b8b4d Add a logo to doc_internal.
5e41ff9b7 Add a hard limit on tcache max size class.
3de19ba40 Eagerly detect double free and sized dealloc bugs for large sizes.
be9548f2b Tcaches: Fix a subtle race condition.
a9aa6f6d0 Fix the alloc_ctx check in free_fastpath.
b971f7c4d Add "default" option to slab sizes.
21b70cb54 Add hpa_central module
1ed7ec369 Emap: Add emap_assert_not_mapped.
2a6ba121b PRNG test: cleanups.
9e6aa77ab PRNG: Remove atomic functionality.
051304717 PRNG: Allow a a range argument of 1.
bdb60a805 Appveyor: don't update msys2 keyring.
025d8c37c Add a script to check for clang-formattedness.
f6bbfc1e9 Add a .clang-format file.
259c5e3e8 psset: Add stats
018b162d6 Add psset: a set of pageslabs.
ed99d300b Flat bitmap: Add longest-range computation.
e03450069 Edata: rename "ranged" bit to "pai".
7ad2f7866 Avoid a -Wundef warning on LG_SLAB_MAXREGS.
40cf71a06 Remove --with-slab-maxregs options from INSTALL.md
36ebb5abe CI support for PPC64LE architecture
1541ffc76 configure: add --with-lg-slab-maxregs configure option.
d243b4ec4 Add PROFILING_INTERNALS.md
09eda2c9b Add unit tests for usize in prof recent records
b549389e4 Correct usize in prof last-N record
202f01d4f Fix szind computation in profiling
866231fc6 Do not repeat reentrancy test in profiling
20f2479ed Do not create size class tables for non-prof builds
8efcdc3f9 Move unbias data to prof_data
5e90fd006 Geom_grow: Don't keep the mutex internal.
c57494879 Geom_grow: Don't take tsdn at init.
ffe552223 Geom_grow: Move in advancing logic.
131b1b533 Rename ecache_grow -> geom_grow.
b399463fb flat_bitmap unit test: Silence a warning.
b0ffa39ca Mallctl stress test: fix a type.
753bbf184 Benchmarks: Also print ns / iter.
7b187360e IO: Support 0-padding for unsigned numbers.
32d467322 Add a mallctl speed stress test.
38867c5c1 Makefile: alphabetize stress/analyze utilities.
ab274a23b Add narenas_ratio.
9e18ae639 Config: safety checks don't imply size checks.
8f9e958e1 Add alignment stress test for rallocx
743021b63 Fix size miscalculation bug in reallocation
eaed1e39b Add sized-delete size-checking functionality.
53084cc5c Safety check: Don't directly abort.
60993697d Prof: Add prof_unbias.
81c2f841e Add a simple utility to detect profiling bias.
e032a1a1d Add a stress test for batch allocation
f6cf5eb38 Add mallctl for batch allocation API
978f830ee Add batch allocation API
c6f59e9bb Add surplus reading API for thread event lookahead
f80546895 Add zero option to arena batch allocation
49e5c2fe7 Add batch allocation from fresh slabs
2bb8060d5 Add empty test and concat for typed list
f28cc2bc8 Extract bin shard selection out of bin locking
ddb8dc4ad FB: Add range iteration support.
ceee82351 Add flat_bitmap.
7fde6ac49 Nbits: Add a couple more interesting sizes.
efeab1f49 bitset test: Pull NBITS_TAB into its own file.
22da83609 bit_util: Add fls_ functions; "find last set".
1ed0288d9 bit_util: Change ffs functions indexing.
786a27b9e CI: Update keyring.
fb347dc61 Verify output space before doing heavy work in mallctl
f5fb4e5a9 Modify mallctl output length when needed
425840204 Corrections for prof_log_start()
e6cb7a1c9 Shorten wait time for peak events
6107857b7 PA->PAC: Move in PAI implementation.
6041aaba9 PA -> PAC: Move in destruction functions.
cbf096b05 Arena: remove redundant bg inactivity check.
471eb5913 PAC: Move in decay rate setting.
6a2774719 PA->PAC: Move in decay functions.
4ee75be3a PA -> PAC: Move in decay_purge enum.
72435b0ab PA->PAC: Make extent.c forget about PA.
dee5d1c42 PA->PAC: Move in extent_sn.
739138234 PA->PAC: Move in stats.
db211eefb PAC: Move in decay.
c81e38999 PAC: Move in ecache_grow.
65803171a PAC: move in emap
7efcb946c PAC: Add an init function.
722652222 PAC: Move in edata_cache accesses.
777b0ba96 Add PAC: Page allocator classic.
1b5f632e0 Introduce PAI: Page allocator interface
3cf19c6e5 atomic: add atomic_load_sub_store
f1f4ec315 Tcache: Tweak nslots_max tuning parameter.
ae541d3fa Edata: Reserve some space for hugepages.
392f645f4 Edata: split up different list linkage uses.
129b72705 Add typed-list module.
00f06c9be enabling mpss on solaris/illumos.
c2e7a0639 No need to intercept prof_dump_header() in tests
f58ebdff7 Generalize prof_cnt_all() for testing
80d18c18c Pass prof dump parameters explicitly in prof_sys
d4259ea53 Simplify signatures for prof dump functions
5d823f3a9 Consolidate struct definitions for prof dump parameters
1f5fe3a3e Pass write callback explicitly in prof_data
4556d3c0c Define structures for prof dump parameters
1c6742e6a Migrate prof dumping to use buffered writer
dad821bb2 Move unwind to prof_sys
d128efcb6 Relocate a few prof utilities to the right modules
4736fb4fc Move file handling logic in prof_data to prof_sys
767a2e179 Move file handling logic in prof to prof_sys
03ae509f3 Create prof_sys module for reading system thread name
adfd9d7b1 Change tsdn to tsd for thread name allocation
841af2b42 Move thread name handling to prof_data module
8118056c0 Expose prof_data testing internals only in prof tests
f43ac8543 Correct prof header macro namings
c8683bee8 Unify printing for prof counts object
5d292b566 Push error handling logic out of core dumping logic
f541871f5 Reduce prof dump buffer size in debug build
354183b10 Define prof dump buffer size centrally
7455813e5 Make dump file writing replaceable in test
21e44c45d Make maps file opening replaceable in test
4bb4037db Extract utility function for opening maps file
f307b2580 Only replace the dump file opening function in test
d8cea8756 Move size inspections to test/analyze
537a4bedb Add a tool to examine random number distributions
d460333ef Improve naming for prof system thread name option
25e43c602 Witness: Make ranks an enum.
092fcac0b Remove unnecessary source files
a795b1932 Remove beginning define in source files
24bbf376c Unify arena flag reading and selection
e128b170a Do not fallback to auto arena when manual arena is requested
95a59d2f7 Unify tcache flag reading and selection
4b0c00848 Unify zero flag reading and setting
2a84f9b8f Unify alignment flag reading and computation
b7858abfc Expose prof testing internal functions
40fa6674a Fix prof timestamp conf reading
7e09a57b3 stress/sizes: Fix an off-by-one issue.
dcfa6fd50 stress/sizes: Add a couple more types.
40672b0b7 Remove duplicate logging in malloc.
4aea74327 High Resolution Timestamps for Profiling
d82a164d0 Add thread.peak.[read|reset] mallctls.
fe7108305 Add peak_t, for tracking allocator net max.
17a64fe91 Add a small program to print data structure sizes.
3e19ebd2e Add lock to protect prof last-N dumping
a835d9cf8 Make prof last-N dumping non-blocking
fc8bc4b5c Increase dump buffer for prof last-N list
264d89d64 Extract restore and async cleanup functions for prof last-N list
857ebd3da Make edata pointer on prof recent record an atomic fence
b8bdea6b2 Fix: prof_recent_alloc_max_ctl_read() does not take tsd
730658f72 Extract alloc/dalloc utility for last-N nodes
035be4486 Separate out dumping for each prof recent record
8da0896b7 Tcache: Make an integer conversion explicit.
cd28e6033 Don't warn on uniform initialization.
6cdac3c57 Tcache: Make flush fractions configurable.
7503b5b33 Stats, CTL: Expose new tcache settings.
ee72bf1cf Tcache: Add tcache gc delay option.
d338dd45d Tcache: Make incremental gc bytes configurable.
ec0b57956 Tcache: Privatize opt_lg_tcache_max default.
10b96f635 Tcache: Remove some unused gc constants.
181093173 Tcache: make slot sizing configurable.
b58dea8d1 Cache bin: expose ncached_max publicly.
634afc412 Tcache: Make size computation configurable.
97b7a9cf7 Add a fill/flush microbenchmark.
33372cbd4 cpu instruction spin wait for arm32/64
27f29e424 LQ_QUANTUM should be 4 on mips64 hardware.
eda9c2858 Edata: zero stack edatas before initializing.
5dead37a9 Allow narenas:default.
dcea2c0f8 Get rid of TSD -> thread event dependency
75dae934a Always initialize TE counters in TSD init
b06dfb9cc Push event handlers to constituent modules
381c97caa Treat postponed prof sample event as new event
abd467493 Extract out per event postponed wait time fetching
f72014d09 Only compute thread event threshold once per trigger
7324c4f85 Break down event init and handler functions
6de77799d Move thread event wait time update to local
733ae918f Extract out per event new wait time fetching
1e2524e15 Do not reset sample wait time when re-initing tdata
855d20f6f Remove outdated comments in thread event
fc052ff72 Migrate counter to use locked int
b543c20a9 Minor update to locked int
f533ab6da Add forking handling for stats
508303077 Add forking handling for prof idump counter
4d970f8bf Add forking handling for counter module
2097e1945 Unify write callback signature
fef9abdcc Cleanup tcache allocation logic
e6cb6919c Consolidate prof inline function headers
d454af90f Remove unused prof_accum field from arena
8be558449 Initialize prof idump counter once rather than once per arena
e10e5059e Make prof_idump_accum() non-inline
039bfd4e3 Do not rollback prof idump counter in arena_prof_promote()
0295aa38a Deduplicate entries in witness error message
f1f8a7549 Let opt.zero propagate to core allocation.
2c09d4349 Add a benchmark of large allocations.
46471ea32 SC: Name the max lookup constant.
79dd0c04e SC: Simplify SC_NPSIZES computation.
fb6cfffd3 Configure: Get rid of LG_QUANTA.
4f8efba82 TSD: Make rtree_ctx a slow-path field.
cd29ebefd Tcache: treat small and large cache bins uniformly
a13fbad37 Tcache: split up fast and slow path data.
7099c6620 Arena: fill in terms of cache_bins.
40e7aed59 TSD: Move in some of the tcache fields.
58a00df23 TSD: Put all fast-path data together.
3589571bf SC: use SC_LG_NGROUP instead of its value.
877af247a QL, QR: Add documentation.
79ae7f921 Rtree: Remove the per-field accessors.
26e9a3103 PA: Simple decay test.
bb6a41852 Emap: Drop szind/slab splitting parameters.
50289750b Extent: Remove szind/slab knowledge.
dc26b3009 Rtree: Clean up compact/non-compact split.
93b99dd14 Extent: Stop passing an edata_cache everywhere.
a4759a191 Ehooks: avoid touching arena_emap_global in tests.
11c47cb13 Extent: Take "bool zero" over "bool *zero".
1a1124462 PA: Take zero as a bool rather than as a bool *.
294b276fc PA: Parameterize emap.  Move emap_global to arena.
f73057727 Eset: Parameterize last globals accesses.
7bb6e2dc0 Eset: take opt_lg_max_active_fit as a parameter.
883ab327c Emap: Move out last edata state touching.
0c96a2f03 Emap: Move out remaining edata modifications.
dfef0df71 Emap: Move edata modification out of emap_remap.
12eb888e5 Edata: Add a ranged bit.
bd4fdf295 Rtree: Pull leaf contents into their own struct.
faec7219b PA: Move in decay initialization.
45671e4a2 PA: Move in retain growth limit setting.
daefde88f PA: Move in mutex stats reading.
07675840a PA: Move in some more internals accesses.
238f3c743 PA: Move in full stats merging.
81c602759 Arena stats: Give it its own "mapped".
506d907e4 PA: Move in basic stats merging.
f29f6090f PA: Add pa_extra.c and put PA forking there.
8164fad40 Stats: Fix edata_cache size merging.
565045ef7 Arena: Make more derived stats non-atomic/locked.
d0c43217b Arena stats: Move retained to PA, use plain ints.
e2cf3fb1a PA: Move in all modifications of mapped.
436789ad9 PA: Make mapped stat atomic.
3c28aa6f1 PA: Move edata_avail stat in, make it non-atomic.
f6bfa3dcc Move extent stats to the PA module.
527dd4cdb PA: Move in nactive counter.
c075fd0bc PA: Minor cleanups and comment fixes.
46a9d7fc0 PA: Move in rest of purging.
2d6eec7b5 PA: Move in decay-all pathway.
65698b7f2 PA: Remove public visibility of some internals.
f012c43be PA: Move in decay_to_limit
103f5feda Move bg thread activity check out of purging core.
3034f4a50 PA: Move in decay_stashed.
aef28b2f8 PA: Move in stash_decayed.
655a09634 Move bg inactivity check out of purge inner loop.
71fc0dc96 PA: Move in remaining page allocation functions.
74958567a PA: have expand take sizes instead of new usize.
5bcc2c2ab PA: Have expand take szind and slab.
0880c2ab9 PA: Have large expands use it.
7be3dea82 PA: Have slab allocations use it.
9f93625c1 PA: Move in arena large allocation functionality.
7624043a4 PA: Add ehook-getting support.
eba35e2e4 Remove extent knowledge of arena.
e77f47a85 Move arena decay getters to PA.
48a2cd6d7 Decay: Add a (mostly stub) test case.
f77cec311 Decay: Take current time as an argument.
bf55e58e6 Rename test/unit/decay -> test/unit/arena_decay.
d1d7e1076 Decay: move in some background_thread accesses.
cdb916ed3 Decay: Add comments for the public API.
8f2193dc8 Decay: Move in arena decay functions.
4d090d23f Decay: Introduce a stub .c file.
7b6288547 Introduce decay module and put decay objects in PA
497836dbc Arena stats: mark edata_avail as derived.
3192d6b77 Extents: Have extent_dalloc_gap take ehooks.
22a0a7b93 Move arena_decay_extent to extent module.
70d12ffa0 PA: Move mapped into pa stats.
6ca918d0c PA: Add a stats comment.
ce8c0d6c0 PA: Move in arena extent_sn counter.
1ada4aef8 PA: Get rid of arena_ind_get calls.
1ad368c8b PA: Move in decay stats.
356aaa7dc Introduce lockedint module.
acd0bf6a2 PA: move in ecache_grow.
32cb7c2f0 PA: Add a stats type.
688fb3eb8 PA: Move in the arena edata_cache.
8433ad84e PA: move in shard initialization.
a24faed56 PA: Move in the ecache_t objects.
585f92505 Move cache index randomization out of extent.
12be9f572 Add a stub PA module -- a page allocator.
c4e9ea8cc Get rid of locks in prof recent test
2deabac07 Get rid of custom iterator for last-N records
a5ddfa7d9 Use ql for prof last-N list
8da6676a0 Don't do reentrant testing in junk tests.
ce17af422 Better structure ql module
4b66297ea Add move constructor to ql module
a62b7ed92 Add emptiness checking to ql module
1dd24ca6d Add rotate functionality to ql module
0dc95a882 Add concat and split functionality to ql module
1ad06aa53 deduplicate insert and delete logic in qr module
c9d56cddf Optimize meld in qr module
0d6d9e858 configure.ac: Put public symbols on one line.
f9aad7a49 Add piping API to buffered writer
09cd79495 Encapsulate buffer allocation failure in buffered writer
a166c2081 Make prof_tctx_t pointer a true prof atomic fence
d936b46d3 Add malloc_conf_2_conf_harder
3b4a03b92 Mac: don't declare system functions as nothrow.
2256ef896 Add option to fetch system thread name on each prof sample
ccdc70a5c Fix: assertion could abort on past failures
b30a5c2f9 Reorganize cpp APIs and suppress unused function warnings
2e5899c12 Stats: Fix tcache_bytes reporting.
a5780598b Remove thread_event_rollback()
ba783b3a0 Remove prof -> thread_event dependency
441d88d1c Rewrite profiling thread event
0dcd57660 Edata cache: atomic fetch-add -> load-store.
99b1291d1 Edata cache: add edata_cache_small_t.
734109d9c Edata cache: add a unit test.
e732344ef Inspect test: Reduce checks when profiling is on.
92485032b Cache bin: improve comments.
d701a085c Fast path: allow low-water mark changes.
397da0386 Cache bin: rewrite to track more state.
fef0b1ffe Cache bin: Remove last internals accesses.
0a2fcfac0 Tcache: Hold cache bin allocation explicitly.
d498a4bb0 Cache bin: Add an emptiness assertion.
6a7aa46ef Cache bin: Add a debug method for init checking.
370c1ea00 Cache bin: Write the unit test in terms of the API
7f5ebd211 Cache bin: set low-water internally.
60113dfe3 Cache bin: Move in initialization code.
44529da85 Cache-bin: Make flush modifications internal
ff6acc6ed Cache bin: simplify names and argument ordering.
e1dcc557d Cache bin: Only take the relevant cache_bin_info_t
1b00d808d cache_bin: Don't let arena see empty position.
d303f3079 cache_bin nflush -> n.
74d36d78e Cache bin: Make ncached_max a query on the info_t.
b66c0973c cache_bin: Don't allow direct internals access.
da68f7329 Move percpu_arena_update.
909c501b0 Cache_bin: Shouldn't know about tcache.
79f1ee2fc Move junking out of arena/tcache code.
b428dceea Config: Warn on void * pointer arithmetic.
22657a5e6 Extents: Silence the "potentially unused" warning.
4a78c6d81 Correct thread event unit test
305b1f6d9 Correction on geometric sampling
6c3491ad3 Tcache: Unify bin flush logic.
9f4fc2738 Ehooks: Fix a build warning.
bc31041ed Cirrus-CI: test on new freebsd releases.
51bd14742 Make use of assert_* in test/unit/thread_event.c
9d2cc3b0f Make use of assert_* in test/unit/prof_recent.c
a88d22ea1 Make use of assert_* in test/unit/inspect.c
0ceb31184 Make use of assert_* in test/unit/buf_writer.c
fa6157938 Add assert_* functionality to tests
21dfa4300 Change assert_* to expect_* in tests
162c2bcf3 Background thread: take base as a parameter.
29436fa05 Break prof and tcache knowledge of b0.
a0c1f4ac5 Rtree: take the base allocator as a parameter.
7013716aa Emap: Take (and propagate) a zeroed parameter.
182192f83 Base: Pull into a single header.
34b7165fd Put szind_t, pszind_t in sz.h.
7e6c8a728 Emap: Standardize naming.
ac50c1e44 Emap: Remove direct access to emap internals.
06e42090f Make jemalloc.c use the emap interface.
f7d9c6c42 Emap: Move in alloc_ctx lookup functionality.
65a54d771 Emap: Move in szind and slab modifications.
9b5d105fc Emap: Move in iealloc.
1d449bd9a Emap: Internal rtree context setting.
08eb1e6c3 Emap: Comments and cleanup
231d1477e Rename emap_split_prepare_t -> emap_prepare_t.
0586a56f3 Emap: Move in merge functionality.
040eac77c Tell edatas their creation arena immediately.
7c7b70206 Emap: Move over metadata splitting logic.
44f5f5360 Emap: Move over deregistration functions.
6513d9d92 Emap: Move over deregistration boundary functions.
9b5ca0b09 Emap: Move in slab interior registration.
d05b61db4 Emap: Move extent boundary registration in.
ca21ce407 Emap: Move in write_acquired from extent.
01f255161 Add emap, for tracking extent locking.
0f686e82a Avoid variable length array with length 0.
68e8ddcaf Add mallctl for dumping last-N profiling records
bc05ecebf Add const qualifier in assert_cmp()
ba0e35411 Rework the bin locking around tcache refill / flush.
7fd22f7b2 Fix Undefined Behavior in hash.h
ca1f08225 Disallow merge across mmap regions to preserve SN / first-fit.
7014f81e1 Add ASSURED_WRITE in mallctl
247688919 Add inspect.c to MSVC filters
9cac3fa8f Encapsulate buffer allocation in buffered writer
bdc08b515 Better naming buffered writer
c6bfe5585 Update the tsd description.
e89652261 Abbreviate thread-event to te.
5e500523a Remove thread_event_boot().
97dd79db6 Implement deallocation events.
536ea6858 NetBSD specific changes: - NetBSD overcommits - When mapping pages, use the maximum of the alignment requested and the   compiled-in PAGE constant which might be greater than the current kernel   pagesize, since we compile binaries with the maximum page size supported   by the architecture (so that they work with all kernels).
974222c62 Add safety check on sdallocx slow / sampled path.
88d9eca84 Enforce page alignment for sampled allocations.
0f552ed67 Don't purge huge extents when decay is off.
38a48e574 Set reentrancy to 1 for tsd_state_purgatory.
88b0e03a4 Implement opt.stats_interval and the _opts options.
d71a145ec Chagne prof_accum_t to counter_accum_t for general purpose.
ea351a7b5 Fix syntax errors in doc for thread.idle.
d92f0175c Introduce NEITHER_READ_NOR_WRITE in ctl.
6a622867c Add "thread.idle" mallctl.
f81341a48 Fallback to unbuffered printing if OOM
cd6e90824 Add stress test for last-N profiling mode
84b28c6a1 Properly handle tdata deletion race
d33120856 Get rid of redundant logic in prof
a72ea0db6 Restructure and correct sleep utility for testing
7b67ed0b5 Get rid of lock overlap in prof_recent_alloc_reset
bd3be8e0b Remove commit parameter to ecache functions.
b8df719d5 No tdata creation for backtracing on dying thread
dab81bd31 Rework and fix the assertions on malloc fastpath.
ad3f3fc56 Fetch time after tctx and only for samples
a5d3dd405 Fix an assertion on extent head state with dss.
2b604a301 Record request size in prof recent entries
40a391408 Define constructor for buffered writer argument
6d8e61690 Make buffered writer an independent module
6b6b4709b Unify buffered writer naming
9a60cf54e Last-N profiling mode
7a27a0594 Delete tdata states used for cleanup
e98ddf798 Fix unlikely condition in arena_prof_info_get()
3fa142cf3 Remove _externs from prof internal header names
112dc36dd Handle log_mtx during forking
ea42174d0 Refactor profiling headers
6342da097 Ehooks: Further optimize default merge case.
f2f2084e7 Ehooks: Assert alloc isn't NULL
e210ccc57 Move extent2 -> extent.
2f4fa8041 Rename extents -> ecache.
56cc56b69 Break extent split dependence on arena.
0aa9769fb Break commit functions' arena dependence
48ec5d435 Break extent_coalesce arena dependence
282a38232 Extent: Break [de]activation's arena dependence.
576d7047a Ecache: Should know its arena_ind.
372042a08 Remove merge dependence on the arena.
439219be7 Remove extent_can_coalesce arena dependency.
9cad5639f Ehooks: remove arena_ind parameter.
57fe99d4b Move relevant index into the ehooks_t itself.
c792f3e4a edata_cache: Remember the associated base_t.
ae23e5f42 Unify extent_alloc_wrapper with the other wrappers.
d8b0b66c6 Put extent_state_t into ecache as well as eset.
98eb40e56 Move delay_coalesce from the eset to the ecache.
bb70df8e5 Extent refactor: Introduce ecache module.
070451624 Ehooks: Add head tracking.
09475bf8a extent_may_dalloc -> ehooks_dalloc_will_fail
785918417 Pull out edata_t caching into its own module.
a7862df61 Rename extent_t to edata_t.
865debda2 Rename extent.h -> edata.h.
a738a66b5 Ehooks: Add some debug zero and addr checks.
4b2e5ee8b Ehooks: Add a "zero" ehook.
d0f187ad3 Arena: Loosen arena_may_have_muzzy restrictions.
ebbb97327 Base: Remove some unnecessary reentrancy guards.
403f2d166 Extents: Split out introspection functionality.
92a511d38 Make extent module hermetic.
e08c581cf Extent: Get rid of extent-specific pre/post reentrancy calls.
39fdc690a Ehooks comments and cleanup.
c8dae890c Extent -> Ehooks: Move over default hooks.
2fe510826 Extent -> Ehooks: Move merge hook.
1fff4d2ee Extent -> Ehooks: Move split hook.
a5b42a1a1 Extent -> Ehooks: Move purge_forced hook.
368baa42e Extent -> Ehooks: Move purge_lazy hook.
f83fdf533 Extent: Clean up a comma
d78fe241a Extent -> Ehooks: Move commit and decommit hooks.
5459ec9da Extent -> Ehooks: Move destroy hook.
bac8e2e5a Extent -> Ehooks: Move dalloc hook.
dc8b4e6e1 Extent -> Ehooks: Move alloc hook.
703fbc0ff Introduce unsafe reentrancy guards.
ae0d8e859 Move extent ehook calls into ehooks
ba8b9ecbc Add ehooks module
837119a94 base_structs.h: Remove some mid-line tabs.
9f6eb0958 Extents: Eagerly initialize extent hooks.
4278f8460 Move extent hook getters/setters to arena.c
9226e1f0d fix opt.thp:never still use THP with base_new
d5031ea82 Allow dallocx and sdallocx after tsd destruction.
4afd709d1 Restructure setters for profiling info
1d01e4c77 Initialization utilities for nstime
dd649c948 Optimize away the tsd_fast() check on fastpath.
1decf958d Fix incorrect usage of cassert.
45836d7fd Pass nstime_t pointer for profiling
7d2bac5a3 Refactor destroy code path for prof_tctx
055478cca Threshold is no longer updated before prof_realloc()
7e3671911 Get rid of old indentation style for prof
dfdd46f6c Refactor prof_tctx_t creation
aa1d71fb7 Rename prof_tctx to alloc_tctx in prof_info_t
5e0b09099 No need to pass usize to prof_tctx_set()
1b1e76acf Disable some spuriously-triggering warnings
a70909b13 Test on all supported release of FreeBSD
5c47a3022 Guard C++ aligned APIs
694537177 Change tsdn to tsd for profiling code path
b55419f9b Restructure profiling
8b2c2a596 Support C++17 over-aligned allocation
9a3c73800 Refactor arena_bin_malloc_hard().
9a7ae3c97 Reduce footprint of bin_t.
cb1a1f4ad Remove the unnecessary alloc_ctx on free_fastpath.
716061710 Add branch hints to free_fastpath.
a787d2f5b Prefer getaffinity() to detect number of CPUs.
04cb7d4d6 Bail out early for muzzy decay.
73510dfd1 Revert "Fix bug in prof_realloc"
3b5eecf10 Fix bug in prof_realloc
e4c36a6f3 Emphasize no modification through thread.allocatedp allowed.
c462753cc Use __forceinline for JEMALLOC_ALWAYS_INLINE on msvc
836d7a7e6 Check for large size first in the uncommon case of malloc.
9c59abe42 Fix a typo in Makefile.
da50d8ce8 Refactor and optimize prof sampling initialization.
bc774a351 Rename tsd->offset_state to tsd->prng_state.
19a51abf3 Avoid arena->offset_state when tsd not available for prng.
d01b425e5 Add -Wimplicit-fallthrough checks if supported
a8b578d53 Remove mallctl test for zero_realloc
43f0ce92d Define general purpose tsd_thread_event_init()
97f93fa0f Pull tcache GC events into thread event handler
198f02e79 Pull prof_accumbytes into thread event handler
152c0ef95 Build a general purpose thread event handler
6924f83cb use SYS_openat when available
de81a4ead Add stats counters for number of zero reallocs
9cfa80594 Realloc: Make behavior of realloc(ptr, 0) configurable.
ee961c231 Merge realloc and rallocx pathways.
bd6e28d6a Guard slabcur fetching in extent_util
4786099a3 Increase column width for global malloc/free rate
05681e387 Optimize cache_bin_alloc_easy for malloc fast path
4fe50bc7d Fix amd64 MSVC warning
4fbbc817c Simplify time setting and getting for prof log
4094b7c03 Limit # of iters of test_bitmap_xfu.
66e07f986 Suppress tdata creation in reentrancy
beb7c16e9 Guard prof_active reset by opt_prof
1df9dd351 Fix je_ prefix issue in test
3d84bd57f Arena: Add helper function arena_get_from_extent.
c97d25575 Eset: Remove temporary declaration.
ce5b128f1 Remove the undefined extent_size_quantize declarations.
821dd53a1 Extent -> Eset: Rename arena members.
e144b21e4 Extent -> Eset: Move fork handling.
77bbb35a9 Extent -> Eset: Move extent fit functions.
1210af9a4 Extent -> Eset: Move insertion and removal.
a42861540 Extents -> Eset: Convert some stats getters.
820f070c6 Move page quantization to sz module.
63d1b7a7a Extents -> Eset: move extents_state_get.
b416b96a3 Extents -> Eset: rename/move extents_init.
e6180fe1b Eset: Add a source file.
4e5e43f22 Rename extents_t -> eset_t.
723ccc6c2 Extents: Split out extent struct.
41187bdfb Extents: Break extent-struct/arena interactions
529cfe2ab Arena: rename arena_structs_b.h -> arena_structs.h
e7cf84a8d Rearrange slab data and constants
d1be488cd Add --with-lg-page=16 to CI.
ac5185f73 Fix tcache bin stack alignment.
b7c7df24b Add max_per_bg_thd stats for per background thread mutexes.
4b76c684b Add "prof.dump_prefix" to override filename prefixes for dumps.
242af439b Rename "prof_dump_seq_mtx" to "prof_dump_filename_mtx".
e06658cb2 check GNU make exists in path
22bc75ee3 Workaround the stringop-overflow check false positives.
93d615180 Pass tsd down to prof_backtrace()
671f120e2 Fix prof_backtrace() reentrancy level
785b84e60 Make cache_bin_sz_t unsigned.
23dc7a7fb Fix index type for cache_bin_alloc_easy.
2abb02ecd Fix MSVC 2015 build, as proposed by @christianaguilera-foundry.
719583f14 Fix large.nflushes in the merged stats.
adce29c88 Optimize for prof_active off
49e6fbce7 Always adjust thread_(de)allocated
57b81c078 Pull thread_(de)allocated out of config_stats
9e031c1d1 Bug fix for prof_active switch
0043e68d4 Track low_water == -1 case explicitly.
937ca1db9 Store ncached_max * ptr_size in tcache_bin_info.
7599c82d4 Redesign the cache bin metadata for fast path.
d2dddfb82 Add hint in the bogus version string.
d6b7995c1 Update INSTALL.md about the default doc build.
e2c758436 Simplify / refactor tcache_dalloc_large.
9c5c2a2c8 Unify the signature of tcache_flush small and large.
28ed9b9a5 Buffer stats printing
eb70fef8c Make compact json format as default
a219cfcda Clear tcache prof_accumbytes in tcache_flush_cache
ad3f7dbfa Buffer prof_log_stop
593484661 Fix large bin index accessed through cache bin descriptor.
22746d3c9 Properly dalloc prof nodes with idalloctm.
8c8466fa6 Add compact json option for emitter
7fc6b1b25 Add buffered writer
39343555d Report stats for tdatas_mtx and prof_dump_mtx
87e2400cb Fix tcaches mutex pre- / post-fork handling.
07ce2434b Refactor profiling
56126d0d2 Refactor prof log
56c8ecffc Correct tsd layout graph

git-subtree-dir: deps/jemalloc
git-subtree-split: 54eaed1d8b56b1aa528be3bdd1877e59c56fa90c
parent 220a0f08
...@@ -2,6 +2,7 @@ ...@@ -2,6 +2,7 @@
#define JEMALLOC_INTERNAL_CACHE_BIN_H #define JEMALLOC_INTERNAL_CACHE_BIN_H
#include "jemalloc/internal/ql.h" #include "jemalloc/internal/ql.h"
#include "jemalloc/internal/sz.h"
/* /*
* The cache_bins are the mechanism that the tcache and the arena use to * The cache_bins are the mechanism that the tcache and the arena use to
...@@ -13,14 +14,38 @@ ...@@ -13,14 +14,38 @@
* of the tcache at all. * of the tcache at all.
*/ */
/*
* The size in bytes of each cache bin stack. We also use this to indicate
* *counts* of individual objects.
*/
typedef uint16_t cache_bin_sz_t;
/* /*
* The count of the number of cached allocations in a bin. We make this signed * Leave a noticeable mark pattern on the cache bin stack boundaries, in case a
* so that negative numbers can encode "invalid" states (e.g. a low water mark * bug starts leaking those. Make it look like the junk pattern but be distinct
* of -1 for a cache that has been depleted). * from it.
*/ */
typedef int32_t cache_bin_sz_t; static const uintptr_t cache_bin_preceding_junk =
(uintptr_t)0x7a7a7a7a7a7a7a7aULL;
/* Note: a7 vs. 7a above -- this tells you which pointer leaked. */
static const uintptr_t cache_bin_trailing_junk =
(uintptr_t)0xa7a7a7a7a7a7a7a7ULL;
/*
* That implies the following value, for the maximum number of items in any
* individual bin. The cache bins track their bounds looking just at the low
* bits of a pointer, compared against a cache_bin_sz_t. So that's
* 1 << (sizeof(cache_bin_sz_t) * 8)
* bytes spread across pointer sized objects to get the maximum.
*/
#define CACHE_BIN_NCACHED_MAX (((size_t)1 << sizeof(cache_bin_sz_t) * 8) \
/ sizeof(void *) - 1)
/*
* This lives inside the cache_bin (for locality reasons), and is initialized
* alongside it, but is otherwise not modified by any cache bin operations.
* It's logically public and maintained by its callers.
*/
typedef struct cache_bin_stats_s cache_bin_stats_t; typedef struct cache_bin_stats_s cache_bin_stats_t;
struct cache_bin_stats_s { struct cache_bin_stats_s {
/* /*
...@@ -36,34 +61,75 @@ struct cache_bin_stats_s { ...@@ -36,34 +61,75 @@ struct cache_bin_stats_s {
*/ */
typedef struct cache_bin_info_s cache_bin_info_t; typedef struct cache_bin_info_s cache_bin_info_t;
struct cache_bin_info_s { struct cache_bin_info_s {
/* Upper limit on ncached. */
cache_bin_sz_t ncached_max; cache_bin_sz_t ncached_max;
}; };
/*
* Responsible for caching allocations associated with a single size.
*
* Several pointers are used to track the stack. To save on metadata bytes,
* only the stack_head is a full sized pointer (which is dereferenced on the
* fastpath), while the others store only the low 16 bits -- this is correct
* because a single stack never takes more space than 2^16 bytes, and at the
* same time only equality checks are performed on the low bits.
*
* (low addr) (high addr)
* |------stashed------|------available------|------cached-----|
* ^ ^ ^ ^
* low_bound(derived) low_bits_full stack_head low_bits_empty
*/
typedef struct cache_bin_s cache_bin_t; typedef struct cache_bin_s cache_bin_t;
struct cache_bin_s { struct cache_bin_s {
/* Min # cached since last GC. */
cache_bin_sz_t low_water;
/* # of cached objects. */
cache_bin_sz_t ncached;
/* /*
* ncached and stats are both modified frequently. Let's keep them * The stack grows down. Whenever the bin is nonempty, the head points
* to an array entry containing a valid allocation. When it is empty,
* the head points to one element past the owned array.
*/
void **stack_head;
/*
* cur_ptr and stats are both modified frequently. Let's keep them
* close so that they have a higher chance of being on the same * close so that they have a higher chance of being on the same
* cacheline, thus less write-backs. * cacheline, thus less write-backs.
*/ */
cache_bin_stats_t tstats; cache_bin_stats_t tstats;
/* /*
* Stack of available objects. * The low bits of the address of the first item in the stack that
* hasn't been used since the last GC, to track the low water mark (min
* # of cached items).
* *
* To make use of adjacent cacheline prefetch, the items in the avail * Since the stack grows down, this is a higher address than
* stack goes to higher address for newer allocations. avail points * low_bits_full.
* just above the available space, which means that
* avail[-ncached, ... -1] are available items and the lowest item will
* be allocated first.
*/ */
void **avail; uint16_t low_bits_low_water;
/*
* The low bits of the value that stack_head will take on when the array
* is full (of cached & stashed items). But remember that stack_head
* always points to a valid item when the array is nonempty -- this is
* in the array.
*
* Recall that since the stack grows down, this is the lowest available
* address in the array for caching. Only adjusted when stashing items.
*/
uint16_t low_bits_full;
/*
* The low bits of the value that stack_head will take on when the array
* is empty.
*
* The stack grows down -- this is one past the highest address in the
* array. Immutable after initialization.
*/
uint16_t low_bits_empty;
}; };
/*
* The cache_bins live inside the tcache, but the arena (by design) isn't
* supposed to know much about tcache internals. To let the arena iterate over
* associated bins, we keep (with the tcache) a linked list of
* cache_bin_array_descriptor_ts that tell the arena how to find the bins.
*/
typedef struct cache_bin_array_descriptor_s cache_bin_array_descriptor_t; typedef struct cache_bin_array_descriptor_s cache_bin_array_descriptor_t;
struct cache_bin_array_descriptor_s { struct cache_bin_array_descriptor_s {
/* /*
...@@ -72,37 +138,214 @@ struct cache_bin_array_descriptor_s { ...@@ -72,37 +138,214 @@ struct cache_bin_array_descriptor_s {
*/ */
ql_elm(cache_bin_array_descriptor_t) link; ql_elm(cache_bin_array_descriptor_t) link;
/* Pointers to the tcache bins. */ /* Pointers to the tcache bins. */
cache_bin_t *bins_small; cache_bin_t *bins;
cache_bin_t *bins_large;
}; };
static inline void static inline void
cache_bin_array_descriptor_init(cache_bin_array_descriptor_t *descriptor, cache_bin_array_descriptor_init(cache_bin_array_descriptor_t *descriptor,
cache_bin_t *bins_small, cache_bin_t *bins_large) { cache_bin_t *bins) {
ql_elm_new(descriptor, link); ql_elm_new(descriptor, link);
descriptor->bins_small = bins_small; descriptor->bins = bins;
descriptor->bins_large = bins_large;
} }
JEMALLOC_ALWAYS_INLINE void * JEMALLOC_ALWAYS_INLINE bool
cache_bin_alloc_easy(cache_bin_t *bin, bool *success) { cache_bin_nonfast_aligned(const void *ptr) {
void *ret; if (!config_uaf_detection) {
return false;
}
/*
* Currently we use alignment to decide which pointer to junk & stash on
* dealloc (for catching use-after-free). In some common cases a
* page-aligned check is needed already (sdalloc w/ config_prof), so we
* are getting it more or less for free -- no added instructions on
* free_fastpath.
*
* Another way of deciding which pointer to sample, is adding another
* thread_event to pick one every N bytes. That also adds no cost on
* the fastpath, however it will tend to pick large allocations which is
* not the desired behavior.
*/
return ((uintptr_t)ptr & san_cache_bin_nonfast_mask) == 0;
}
/* Returns ncached_max: Upper limit on ncached. */
static inline cache_bin_sz_t
cache_bin_info_ncached_max(cache_bin_info_t *info) {
return info->ncached_max;
}
bin->ncached--; /*
* Internal.
*
* Asserts that the pointer associated with earlier is <= the one associated
* with later.
*/
static inline void
cache_bin_assert_earlier(cache_bin_t *bin, uint16_t earlier, uint16_t later) {
if (earlier > later) {
assert(bin->low_bits_full > bin->low_bits_empty);
}
}
/*
* Internal.
*
* Does difference calculations that handle wraparound correctly. Earlier must
* be associated with the position earlier in memory.
*/
static inline uint16_t
cache_bin_diff(cache_bin_t *bin, uint16_t earlier, uint16_t later, bool racy) {
/* /*
* Check for both bin->ncached == 0 and ncached < low_water * When it's racy, bin->low_bits_full can be modified concurrently. It
* in a single branch. * can cross the uint16_t max value and become less than
* bin->low_bits_empty at the time of the check.
*/ */
if (unlikely(bin->ncached <= bin->low_water)) { if (!racy) {
bin->low_water = bin->ncached; cache_bin_assert_earlier(bin, earlier, later);
if (bin->ncached == -1) {
bin->ncached = 0;
*success = false;
return NULL;
} }
return later - earlier;
}
/*
* Number of items currently cached in the bin, without checking ncached_max.
* We require specifying whether or not the request is racy or not (i.e. whether
* or not concurrent modifications are possible).
*/
static inline cache_bin_sz_t
cache_bin_ncached_get_internal(cache_bin_t *bin, bool racy) {
cache_bin_sz_t diff = cache_bin_diff(bin,
(uint16_t)(uintptr_t)bin->stack_head, bin->low_bits_empty, racy);
cache_bin_sz_t n = diff / sizeof(void *);
/*
* We have undefined behavior here; if this function is called from the
* arena stats updating code, then stack_head could change from the
* first line to the next one. Morally, these loads should be atomic,
* but compilers won't currently generate comparisons with in-memory
* operands against atomics, and these variables get accessed on the
* fast paths. This should still be "safe" in the sense of generating
* the correct assembly for the foreseeable future, though.
*/
assert(n == 0 || *(bin->stack_head) != NULL || racy);
return n;
}
/*
* Number of items currently cached in the bin, with checking ncached_max. The
* caller must know that no concurrent modification of the cache_bin is
* possible.
*/
static inline cache_bin_sz_t
cache_bin_ncached_get_local(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin,
/* racy */ false);
assert(n <= cache_bin_info_ncached_max(info));
return n;
}
/*
* Internal.
*
* A pointer to the position one past the end of the backing array.
*
* Do not call if racy, because both 'bin->stack_head' and 'bin->low_bits_full'
* are subject to concurrent modifications.
*/
static inline void **
cache_bin_empty_position_get(cache_bin_t *bin) {
cache_bin_sz_t diff = cache_bin_diff(bin,
(uint16_t)(uintptr_t)bin->stack_head, bin->low_bits_empty,
/* racy */ false);
uintptr_t empty_bits = (uintptr_t)bin->stack_head + diff;
void **ret = (void **)empty_bits;
assert(ret >= bin->stack_head);
return ret;
}
/*
* Internal.
*
* Calculates low bits of the lower bound of the usable cache bin's range (see
* cache_bin_t visual representation above).
*
* No values are concurrently modified, so should be safe to read in a
* multithreaded environment. Currently concurrent access happens only during
* arena statistics collection.
*/
static inline uint16_t
cache_bin_low_bits_low_bound_get(cache_bin_t *bin, cache_bin_info_t *info) {
return (uint16_t)bin->low_bits_empty -
info->ncached_max * sizeof(void *);
}
/*
* Internal.
*
* A pointer to the position with the lowest address of the backing array.
*/
static inline void **
cache_bin_low_bound_get(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_sz_t ncached_max = cache_bin_info_ncached_max(info);
void **ret = cache_bin_empty_position_get(bin) - ncached_max;
assert(ret <= bin->stack_head);
return ret;
}
/*
* As the name implies. This is important since it's not correct to try to
* batch fill a nonempty cache bin.
*/
static inline void
cache_bin_assert_empty(cache_bin_t *bin, cache_bin_info_t *info) {
assert(cache_bin_ncached_get_local(bin, info) == 0);
assert(cache_bin_empty_position_get(bin) == bin->stack_head);
}
/*
* Get low water, but without any of the correctness checking we do for the
* caller-usable version, if we are temporarily breaking invariants (like
* ncached >= low_water during flush).
*/
static inline cache_bin_sz_t
cache_bin_low_water_get_internal(cache_bin_t *bin) {
return cache_bin_diff(bin, bin->low_bits_low_water,
bin->low_bits_empty, /* racy */ false) / sizeof(void *);
}
/* Returns the numeric value of low water in [0, ncached]. */
static inline cache_bin_sz_t
cache_bin_low_water_get(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_sz_t low_water = cache_bin_low_water_get_internal(bin);
assert(low_water <= cache_bin_info_ncached_max(info));
assert(low_water <= cache_bin_ncached_get_local(bin, info));
cache_bin_assert_earlier(bin, (uint16_t)(uintptr_t)bin->stack_head,
bin->low_bits_low_water);
return low_water;
}
/*
* Indicates that the current cache bin position should be the low water mark
* going forward.
*/
static inline void
cache_bin_low_water_set(cache_bin_t *bin) {
bin->low_bits_low_water = (uint16_t)(uintptr_t)bin->stack_head;
}
static inline void
cache_bin_low_water_adjust(cache_bin_t *bin) {
if (cache_bin_ncached_get_internal(bin, /* racy */ false)
< cache_bin_low_water_get_internal(bin)) {
cache_bin_low_water_set(bin);
} }
}
JEMALLOC_ALWAYS_INLINE void *
cache_bin_alloc_impl(cache_bin_t *bin, bool *success, bool adjust_low_water) {
/* /*
* success (instead of ret) should be checked upon the return of this * success (instead of ret) should be checked upon the return of this
* function. We avoid checking (ret == NULL) because there is never a * function. We avoid checking (ret == NULL) because there is never a
...@@ -110,22 +353,318 @@ cache_bin_alloc_easy(cache_bin_t *bin, bool *success) { ...@@ -110,22 +353,318 @@ cache_bin_alloc_easy(cache_bin_t *bin, bool *success) {
* and eagerly checking ret would cause pipeline stall (waiting for the * and eagerly checking ret would cause pipeline stall (waiting for the
* cacheline). * cacheline).
*/ */
*success = true;
ret = *(bin->avail - (bin->ncached + 1));
/*
* This may read from the empty position; however the loaded value won't
* be used. It's safe because the stack has one more slot reserved.
*/
void *ret = *bin->stack_head;
uint16_t low_bits = (uint16_t)(uintptr_t)bin->stack_head;
void **new_head = bin->stack_head + 1;
/*
* Note that the low water mark is at most empty; if we pass this check,
* we know we're non-empty.
*/
if (likely(low_bits != bin->low_bits_low_water)) {
bin->stack_head = new_head;
*success = true;
return ret; return ret;
}
if (!adjust_low_water) {
*success = false;
return NULL;
}
/*
* In the fast-path case where we call alloc_easy and then alloc, the
* previous checking and computation is optimized away -- we didn't
* actually commit any of our operations.
*/
if (likely(low_bits != bin->low_bits_empty)) {
bin->stack_head = new_head;
bin->low_bits_low_water = (uint16_t)(uintptr_t)new_head;
*success = true;
return ret;
}
*success = false;
return NULL;
}
/*
* Allocate an item out of the bin, failing if we're at the low-water mark.
*/
JEMALLOC_ALWAYS_INLINE void *
cache_bin_alloc_easy(cache_bin_t *bin, bool *success) {
/* We don't look at info if we're not adjusting low-water. */
return cache_bin_alloc_impl(bin, success, false);
}
/*
* Allocate an item out of the bin, even if we're currently at the low-water
* mark (and failing only if the bin is empty).
*/
JEMALLOC_ALWAYS_INLINE void *
cache_bin_alloc(cache_bin_t *bin, bool *success) {
return cache_bin_alloc_impl(bin, success, true);
}
JEMALLOC_ALWAYS_INLINE cache_bin_sz_t
cache_bin_alloc_batch(cache_bin_t *bin, size_t num, void **out) {
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin,
/* racy */ false);
if (n > num) {
n = (cache_bin_sz_t)num;
}
memcpy(out, bin->stack_head, n * sizeof(void *));
bin->stack_head += n;
cache_bin_low_water_adjust(bin);
return n;
}
JEMALLOC_ALWAYS_INLINE bool
cache_bin_full(cache_bin_t *bin) {
return ((uint16_t)(uintptr_t)bin->stack_head == bin->low_bits_full);
}
/*
* Free an object into the given bin. Fails only if the bin is full.
*/
JEMALLOC_ALWAYS_INLINE bool
cache_bin_dalloc_easy(cache_bin_t *bin, void *ptr) {
if (unlikely(cache_bin_full(bin))) {
return false;
}
bin->stack_head--;
*bin->stack_head = ptr;
cache_bin_assert_earlier(bin, bin->low_bits_full,
(uint16_t)(uintptr_t)bin->stack_head);
return true;
} }
/* Returns false if failed to stash (i.e. bin is full). */
JEMALLOC_ALWAYS_INLINE bool JEMALLOC_ALWAYS_INLINE bool
cache_bin_dalloc_easy(cache_bin_t *bin, cache_bin_info_t *bin_info, void *ptr) { cache_bin_stash(cache_bin_t *bin, void *ptr) {
if (unlikely(bin->ncached == bin_info->ncached_max)) { if (cache_bin_full(bin)) {
return false; return false;
} }
assert(bin->ncached < bin_info->ncached_max);
bin->ncached++; /* Stash at the full position, in the [full, head) range. */
*(bin->avail - bin->ncached) = ptr; uint16_t low_bits_head = (uint16_t)(uintptr_t)bin->stack_head;
/* Wraparound handled as well. */
uint16_t diff = cache_bin_diff(bin, bin->low_bits_full, low_bits_head,
/* racy */ false);
*(void **)((uintptr_t)bin->stack_head - diff) = ptr;
assert(!cache_bin_full(bin));
bin->low_bits_full += sizeof(void *);
cache_bin_assert_earlier(bin, bin->low_bits_full, low_bits_head);
return true; return true;
} }
/*
* Get the number of stashed pointers.
*
* When called from a thread not owning the TLS (i.e. racy = true), it's
* important to keep in mind that 'bin->stack_head' and 'bin->low_bits_full' can
* be modified concurrently and almost none assertions about their values can be
* made.
*/
JEMALLOC_ALWAYS_INLINE cache_bin_sz_t
cache_bin_nstashed_get_internal(cache_bin_t *bin, cache_bin_info_t *info,
bool racy) {
cache_bin_sz_t ncached_max = cache_bin_info_ncached_max(info);
uint16_t low_bits_low_bound = cache_bin_low_bits_low_bound_get(bin,
info);
cache_bin_sz_t n = cache_bin_diff(bin, low_bits_low_bound,
bin->low_bits_full, racy) / sizeof(void *);
assert(n <= ncached_max);
if (!racy) {
/* Below are for assertions only. */
void **low_bound = cache_bin_low_bound_get(bin, info);
assert((uint16_t)(uintptr_t)low_bound == low_bits_low_bound);
void *stashed = *(low_bound + n - 1);
bool aligned = cache_bin_nonfast_aligned(stashed);
#ifdef JEMALLOC_JET
/* Allow arbitrary pointers to be stashed in tests. */
aligned = true;
#endif
assert(n == 0 || (stashed != NULL && aligned));
}
return n;
}
JEMALLOC_ALWAYS_INLINE cache_bin_sz_t
cache_bin_nstashed_get_local(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_sz_t n = cache_bin_nstashed_get_internal(bin, info,
/* racy */ false);
assert(n <= cache_bin_info_ncached_max(info));
return n;
}
/*
* Obtain a racy view of the number of items currently in the cache bin, in the
* presence of possible concurrent modifications.
*/
static inline void
cache_bin_nitems_get_remote(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_sz_t *ncached, cache_bin_sz_t *nstashed) {
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin, /* racy */ true);
assert(n <= cache_bin_info_ncached_max(info));
*ncached = n;
n = cache_bin_nstashed_get_internal(bin, info, /* racy */ true);
assert(n <= cache_bin_info_ncached_max(info));
*nstashed = n;
/* Note that cannot assert ncached + nstashed <= ncached_max (racy). */
}
/*
* Filling and flushing are done in batch, on arrays of void *s. For filling,
* the arrays go forward, and can be accessed with ordinary array arithmetic.
* For flushing, we work from the end backwards, and so need to use special
* accessors that invert the usual ordering.
*
* This is important for maintaining first-fit; the arena code fills with
* earliest objects first, and so those are the ones we should return first for
* cache_bin_alloc calls. When flushing, we should flush the objects that we
* wish to return later; those at the end of the array. This is better for the
* first-fit heuristic as well as for cache locality; the most recently freed
* objects are the ones most likely to still be in cache.
*
* This all sounds very hand-wavey and theoretical, but reverting the ordering
* on one or the other pathway leads to measurable slowdowns.
*/
typedef struct cache_bin_ptr_array_s cache_bin_ptr_array_t;
struct cache_bin_ptr_array_s {
cache_bin_sz_t n;
void **ptr;
};
/*
* Declare a cache_bin_ptr_array_t sufficient for nval items.
*
* In the current implementation, this could be just part of a
* cache_bin_ptr_array_init_... call, since we reuse the cache bin stack memory.
* Indirecting behind a macro, though, means experimenting with linked-list
* representations is easy (since they'll require an alloca in the calling
* frame).
*/
#define CACHE_BIN_PTR_ARRAY_DECLARE(name, nval) \
cache_bin_ptr_array_t name; \
name.n = (nval)
/*
* Start a fill. The bin must be empty, and This must be followed by a
* finish_fill call before doing any alloc/dalloc operations on the bin.
*/
static inline void
cache_bin_init_ptr_array_for_fill(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nfill) {
cache_bin_assert_empty(bin, info);
arr->ptr = cache_bin_empty_position_get(bin) - nfill;
}
/*
* While nfill in cache_bin_init_ptr_array_for_fill is the number we *intend* to
* fill, nfilled here is the number we actually filled (which may be less, in
* case of OOM.
*/
static inline void
cache_bin_finish_fill(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nfilled) {
cache_bin_assert_empty(bin, info);
void **empty_position = cache_bin_empty_position_get(bin);
if (nfilled < arr->n) {
memmove(empty_position - nfilled, empty_position - arr->n,
nfilled * sizeof(void *));
}
bin->stack_head = empty_position - nfilled;
}
/*
* Same deal, but with flush. Unlike fill (which can fail), the user must flush
* everything we give them.
*/
static inline void
cache_bin_init_ptr_array_for_flush(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nflush) {
arr->ptr = cache_bin_empty_position_get(bin) - nflush;
assert(cache_bin_ncached_get_local(bin, info) == 0
|| *arr->ptr != NULL);
}
static inline void
cache_bin_finish_flush(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nflushed) {
unsigned rem = cache_bin_ncached_get_local(bin, info) - nflushed;
memmove(bin->stack_head + nflushed, bin->stack_head,
rem * sizeof(void *));
bin->stack_head = bin->stack_head + nflushed;
cache_bin_low_water_adjust(bin);
}
static inline void
cache_bin_init_ptr_array_for_stashed(cache_bin_t *bin, szind_t binind,
cache_bin_info_t *info, cache_bin_ptr_array_t *arr,
cache_bin_sz_t nstashed) {
assert(nstashed > 0);
assert(cache_bin_nstashed_get_local(bin, info) == nstashed);
void **low_bound = cache_bin_low_bound_get(bin, info);
arr->ptr = low_bound;
assert(*arr->ptr != NULL);
}
static inline void
cache_bin_finish_flush_stashed(cache_bin_t *bin, cache_bin_info_t *info) {
void **low_bound = cache_bin_low_bound_get(bin, info);
/* Reset the bin local full position. */
bin->low_bits_full = (uint16_t)(uintptr_t)low_bound;
assert(cache_bin_nstashed_get_local(bin, info) == 0);
}
/*
* Initialize a cache_bin_info to represent up to the given number of items in
* the cache_bins it is associated with.
*/
void cache_bin_info_init(cache_bin_info_t *bin_info,
cache_bin_sz_t ncached_max);
/*
* Given an array of initialized cache_bin_info_ts, determine how big an
* allocation is required to initialize a full set of cache_bin_ts.
*/
void cache_bin_info_compute_alloc(cache_bin_info_t *infos, szind_t ninfos,
size_t *size, size_t *alignment);
/*
* Actually initialize some cache bins. Callers should allocate the backing
* memory indicated by a call to cache_bin_compute_alloc. They should then
* preincrement, call init once for each bin and info, and then call
* cache_bin_postincrement. *alloc_cur will then point immediately past the end
* of the allocation.
*/
void cache_bin_preincrement(cache_bin_info_t *infos, szind_t ninfos,
void *alloc, size_t *cur_offset);
void cache_bin_postincrement(cache_bin_info_t *infos, szind_t ninfos,
void *alloc, size_t *cur_offset);
void cache_bin_init(cache_bin_t *bin, cache_bin_info_t *info, void *alloc,
size_t *cur_offset);
/*
* If a cache bin was zero initialized (either because it lives in static or
* thread-local storage, or was memset to 0), this function indicates whether or
* not cache_bin_init was called on it.
*/
bool cache_bin_still_zero_initialized(cache_bin_t *bin);
#endif /* JEMALLOC_INTERNAL_CACHE_BIN_H */ #endif /* JEMALLOC_INTERNAL_CACHE_BIN_H */
#ifndef JEMALLOC_INTERNAL_COUNTER_H
#define JEMALLOC_INTERNAL_COUNTER_H
#include "jemalloc/internal/mutex.h"
typedef struct counter_accum_s {
LOCKEDINT_MTX_DECLARE(mtx)
locked_u64_t accumbytes;
uint64_t interval;
} counter_accum_t;
JEMALLOC_ALWAYS_INLINE bool
counter_accum(tsdn_t *tsdn, counter_accum_t *counter, uint64_t bytes) {
uint64_t interval = counter->interval;
assert(interval > 0);
LOCKEDINT_MTX_LOCK(tsdn, counter->mtx);
/*
* If the event moves fast enough (and/or if the event handling is slow
* enough), extreme overflow can cause counter trigger coalescing.
* This is an intentional mechanism that avoids rate-limiting
* allocation.
*/
bool overflow = locked_inc_mod_u64(tsdn, LOCKEDINT_MTX(counter->mtx),
&counter->accumbytes, bytes, interval);
LOCKEDINT_MTX_UNLOCK(tsdn, counter->mtx);
return overflow;
}
bool counter_accum_init(counter_accum_t *counter, uint64_t interval);
void counter_prefork(tsdn_t *tsdn, counter_accum_t *counter);
void counter_postfork_parent(tsdn_t *tsdn, counter_accum_t *counter);
void counter_postfork_child(tsdn_t *tsdn, counter_accum_t *counter);
#endif /* JEMALLOC_INTERNAL_COUNTER_H */
...@@ -42,9 +42,11 @@ typedef struct ctl_arena_stats_s { ...@@ -42,9 +42,11 @@ typedef struct ctl_arena_stats_s {
uint64_t nfills_small; uint64_t nfills_small;
uint64_t nflushes_small; uint64_t nflushes_small;
bin_stats_t bstats[SC_NBINS]; bin_stats_data_t bstats[SC_NBINS];
arena_stats_large_t lstats[SC_NSIZES - SC_NBINS]; arena_stats_large_t lstats[SC_NSIZES - SC_NBINS];
arena_stats_extents_t estats[SC_NPSIZES]; pac_estats_t estats[SC_NPSIZES];
hpa_shard_stats_t hpastats;
sec_stats_t secstats;
} ctl_arena_stats_t; } ctl_arena_stats_t;
typedef struct ctl_stats_s { typedef struct ctl_stats_s {
...@@ -96,13 +98,17 @@ typedef struct ctl_arenas_s { ...@@ -96,13 +98,17 @@ typedef struct ctl_arenas_s {
int ctl_byname(tsd_t *tsd, const char *name, void *oldp, size_t *oldlenp, int ctl_byname(tsd_t *tsd, const char *name, void *oldp, size_t *oldlenp,
void *newp, size_t newlen); void *newp, size_t newlen);
int ctl_nametomib(tsd_t *tsd, const char *name, size_t *mibp, size_t *miblenp); int ctl_nametomib(tsd_t *tsd, const char *name, size_t *mibp, size_t *miblenp);
int ctl_bymib(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp, int ctl_bymib(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
size_t *oldlenp, void *newp, size_t newlen); size_t *oldlenp, void *newp, size_t newlen);
int ctl_mibnametomib(tsd_t *tsd, size_t *mib, size_t miblen, const char *name,
size_t *miblenp);
int ctl_bymibname(tsd_t *tsd, size_t *mib, size_t miblen, const char *name,
size_t *miblenp, void *oldp, size_t *oldlenp, void *newp, size_t newlen);
bool ctl_boot(void); bool ctl_boot(void);
void ctl_prefork(tsdn_t *tsdn); void ctl_prefork(tsdn_t *tsdn);
void ctl_postfork_parent(tsdn_t *tsdn); void ctl_postfork_parent(tsdn_t *tsdn);
void ctl_postfork_child(tsdn_t *tsdn); void ctl_postfork_child(tsdn_t *tsdn);
void ctl_mtx_assert_held(tsdn_t *tsdn);
#define xmallctl(name, oldp, oldlenp, newp, newlen) do { \ #define xmallctl(name, oldp, oldlenp, newp, newlen) do { \
if (je_mallctl(name, oldp, oldlenp, newp, newlen) \ if (je_mallctl(name, oldp, oldlenp, newp, newlen) \
...@@ -131,4 +137,23 @@ void ctl_postfork_child(tsdn_t *tsdn); ...@@ -131,4 +137,23 @@ void ctl_postfork_child(tsdn_t *tsdn);
} \ } \
} while (0) } while (0)
#define xmallctlmibnametomib(mib, miblen, name, miblenp) do { \
if (ctl_mibnametomib(tsd_fetch(), mib, miblen, name, miblenp) \
!= 0) { \
malloc_write( \
"<jemalloc>: Failure in ctl_mibnametomib()\n"); \
abort(); \
} \
} while (0)
#define xmallctlbymibname(mib, miblen, name, miblenp, oldp, oldlenp, \
newp, newlen) do { \
if (ctl_bymibname(tsd_fetch(), mib, miblen, name, miblenp, \
oldp, oldlenp, newp, newlen) != 0) { \
malloc_write( \
"<jemalloc>: Failure in ctl_bymibname()\n"); \
abort(); \
} \
} while (0)
#endif /* JEMALLOC_INTERNAL_CTL_H */ #endif /* JEMALLOC_INTERNAL_CTL_H */
#ifndef JEMALLOC_INTERNAL_DECAY_H
#define JEMALLOC_INTERNAL_DECAY_H
#include "jemalloc/internal/smoothstep.h"
#define DECAY_UNBOUNDED_TIME_TO_PURGE ((uint64_t)-1)
/*
* The decay_t computes the number of pages we should purge at any given time.
* Page allocators inform a decay object when pages enter a decay-able state
* (i.e. dirty or muzzy), and query it to determine how many pages should be
* purged at any given time.
*
* This is mostly a single-threaded data structure and doesn't care about
* synchronization at all; it's the caller's responsibility to manage their
* synchronization on their own. There are two exceptions:
* 1) It's OK to racily call decay_ms_read (i.e. just the simplest state query).
* 2) The mtx and purging fields live (and are initialized) here, but are
* logically owned by the page allocator. This is just a convenience (since
* those fields would be duplicated for both the dirty and muzzy states
* otherwise).
*/
typedef struct decay_s decay_t;
struct decay_s {
/* Synchronizes all non-atomic fields. */
malloc_mutex_t mtx;
/*
* True if a thread is currently purging the extents associated with
* this decay structure.
*/
bool purging;
/*
* Approximate time in milliseconds from the creation of a set of unused
* dirty pages until an equivalent set of unused dirty pages is purged
* and/or reused.
*/
atomic_zd_t time_ms;
/* time / SMOOTHSTEP_NSTEPS. */
nstime_t interval;
/*
* Time at which the current decay interval logically started. We do
* not actually advance to a new epoch until sometime after it starts
* because of scheduling and computation delays, and it is even possible
* to completely skip epochs. In all cases, during epoch advancement we
* merge all relevant activity into the most recently recorded epoch.
*/
nstime_t epoch;
/* Deadline randomness generator. */
uint64_t jitter_state;
/*
* Deadline for current epoch. This is the sum of interval and per
* epoch jitter which is a uniform random variable in [0..interval).
* Epochs always advance by precise multiples of interval, but we
* randomize the deadline to reduce the likelihood of arenas purging in
* lockstep.
*/
nstime_t deadline;
/*
* The number of pages we cap ourselves at in the current epoch, per
* decay policies. Updated on an epoch change. After an epoch change,
* the caller should take steps to try to purge down to this amount.
*/
size_t npages_limit;
/*
* Number of unpurged pages at beginning of current epoch. During epoch
* advancement we use the delta between arena->decay_*.nunpurged and
* ecache_npages_get(&arena->ecache_*) to determine how many dirty pages,
* if any, were generated.
*/
size_t nunpurged;
/*
* Trailing log of how many unused dirty pages were generated during
* each of the past SMOOTHSTEP_NSTEPS decay epochs, where the last
* element is the most recent epoch. Corresponding epoch times are
* relative to epoch.
*
* Updated only on epoch advance, triggered by
* decay_maybe_advance_epoch, below.
*/
size_t backlog[SMOOTHSTEP_NSTEPS];
/* Peak number of pages in associated extents. Used for debug only. */
uint64_t ceil_npages;
};
/*
* The current decay time setting. This is the only public access to a decay_t
* that's allowed without holding mtx.
*/
static inline ssize_t
decay_ms_read(const decay_t *decay) {
return atomic_load_zd(&decay->time_ms, ATOMIC_RELAXED);
}
/*
* See the comment on the struct field -- the limit on pages we should allow in
* this decay state this epoch.
*/
static inline size_t
decay_npages_limit_get(const decay_t *decay) {
return decay->npages_limit;
}
/* How many unused dirty pages were generated during the last epoch. */
static inline size_t
decay_epoch_npages_delta(const decay_t *decay) {
return decay->backlog[SMOOTHSTEP_NSTEPS - 1];
}
/*
* Current epoch duration, in nanoseconds. Given that new epochs are started
* somewhat haphazardly, this is not necessarily exactly the time between any
* two calls to decay_maybe_advance_epoch; see the comments on fields in the
* decay_t.
*/
static inline uint64_t
decay_epoch_duration_ns(const decay_t *decay) {
return nstime_ns(&decay->interval);
}
static inline bool
decay_immediately(const decay_t *decay) {
ssize_t decay_ms = decay_ms_read(decay);
return decay_ms == 0;
}
static inline bool
decay_disabled(const decay_t *decay) {
ssize_t decay_ms = decay_ms_read(decay);
return decay_ms < 0;
}
/* Returns true if decay is enabled and done gradually. */
static inline bool
decay_gradually(const decay_t *decay) {
ssize_t decay_ms = decay_ms_read(decay);
return decay_ms > 0;
}
/*
* Returns true if the passed in decay time setting is valid.
* < -1 : invalid
* -1 : never decay
* 0 : decay immediately
* > 0 : some positive decay time, up to a maximum allowed value of
* NSTIME_SEC_MAX * 1000, which corresponds to decaying somewhere in the early
* 27th century. By that time, we expect to have implemented alternate purging
* strategies.
*/
bool decay_ms_valid(ssize_t decay_ms);
/*
* As a precondition, the decay_t must be zeroed out (as if with memset).
*
* Returns true on error.
*/
bool decay_init(decay_t *decay, nstime_t *cur_time, ssize_t decay_ms);
/*
* Given an already-initialized decay_t, reinitialize it with the given decay
* time. The decay_t must have previously been initialized (and should not then
* be zeroed).
*/
void decay_reinit(decay_t *decay, nstime_t *cur_time, ssize_t decay_ms);
/*
* Compute how many of 'npages_new' pages we would need to purge in 'time'.
*/
uint64_t decay_npages_purge_in(decay_t *decay, nstime_t *time,
size_t npages_new);
/* Returns true if the epoch advanced and there are pages to purge. */
bool decay_maybe_advance_epoch(decay_t *decay, nstime_t *new_time,
size_t current_npages);
/*
* Calculates wait time until a number of pages in the interval
* [0.5 * npages_threshold .. 1.5 * npages_threshold] should be purged.
*
* Returns number of nanoseconds or DECAY_UNBOUNDED_TIME_TO_PURGE in case of
* indefinite wait.
*/
uint64_t decay_ns_until_purge(decay_t *decay, size_t npages_current,
uint64_t npages_threshold);
#endif /* JEMALLOC_INTERNAL_DECAY_H */
#ifndef JEMALLOC_INTERNAL_ECACHE_H
#define JEMALLOC_INTERNAL_ECACHE_H
#include "jemalloc/internal/eset.h"
#include "jemalloc/internal/san.h"
#include "jemalloc/internal/mutex.h"
typedef struct ecache_s ecache_t;
struct ecache_s {
malloc_mutex_t mtx;
eset_t eset;
eset_t guarded_eset;
/* All stored extents must be in the same state. */
extent_state_t state;
/* The index of the ehooks the ecache is associated with. */
unsigned ind;
/*
* If true, delay coalescing until eviction; otherwise coalesce during
* deallocation.
*/
bool delay_coalesce;
};
static inline size_t
ecache_npages_get(ecache_t *ecache) {
return eset_npages_get(&ecache->eset) +
eset_npages_get(&ecache->guarded_eset);
}
/* Get the number of extents in the given page size index. */
static inline size_t
ecache_nextents_get(ecache_t *ecache, pszind_t ind) {
return eset_nextents_get(&ecache->eset, ind) +
eset_nextents_get(&ecache->guarded_eset, ind);
}
/* Get the sum total bytes of the extents in the given page size index. */
static inline size_t
ecache_nbytes_get(ecache_t *ecache, pszind_t ind) {
return eset_nbytes_get(&ecache->eset, ind) +
eset_nbytes_get(&ecache->guarded_eset, ind);
}
static inline unsigned
ecache_ind_get(ecache_t *ecache) {
return ecache->ind;
}
bool ecache_init(tsdn_t *tsdn, ecache_t *ecache, extent_state_t state,
unsigned ind, bool delay_coalesce);
void ecache_prefork(tsdn_t *tsdn, ecache_t *ecache);
void ecache_postfork_parent(tsdn_t *tsdn, ecache_t *ecache);
void ecache_postfork_child(tsdn_t *tsdn, ecache_t *ecache);
#endif /* JEMALLOC_INTERNAL_ECACHE_H */
#ifndef JEMALLOC_INTERNAL_EDATA_H
#define JEMALLOC_INTERNAL_EDATA_H
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/bin_info.h"
#include "jemalloc/internal/bit_util.h"
#include "jemalloc/internal/hpdata.h"
#include "jemalloc/internal/nstime.h"
#include "jemalloc/internal/ph.h"
#include "jemalloc/internal/ql.h"
#include "jemalloc/internal/sc.h"
#include "jemalloc/internal/slab_data.h"
#include "jemalloc/internal/sz.h"
#include "jemalloc/internal/typed_list.h"
/*
* sizeof(edata_t) is 128 bytes on 64-bit architectures. Ensure the alignment
* to free up the low bits in the rtree leaf.
*/
#define EDATA_ALIGNMENT 128
enum extent_state_e {
extent_state_active = 0,
extent_state_dirty = 1,
extent_state_muzzy = 2,
extent_state_retained = 3,
extent_state_transition = 4, /* States below are intermediate. */
extent_state_merging = 5,
extent_state_max = 5 /* Sanity checking only. */
};
typedef enum extent_state_e extent_state_t;
enum extent_head_state_e {
EXTENT_NOT_HEAD,
EXTENT_IS_HEAD /* See comments in ehooks_default_merge_impl(). */
};
typedef enum extent_head_state_e extent_head_state_t;
/*
* Which implementation of the page allocator interface, (PAI, defined in
* pai.h) owns the given extent?
*/
enum extent_pai_e {
EXTENT_PAI_PAC = 0,
EXTENT_PAI_HPA = 1
};
typedef enum extent_pai_e extent_pai_t;
struct e_prof_info_s {
/* Time when this was allocated. */
nstime_t e_prof_alloc_time;
/* Allocation request size. */
size_t e_prof_alloc_size;
/* Points to a prof_tctx_t. */
atomic_p_t e_prof_tctx;
/*
* Points to a prof_recent_t for the allocation; NULL
* means the recent allocation record no longer exists.
* Protected by prof_recent_alloc_mtx.
*/
atomic_p_t e_prof_recent_alloc;
};
typedef struct e_prof_info_s e_prof_info_t;
/*
* The information about a particular edata that lives in an emap. Space is
* more precious there (the information, plus the edata pointer, has to live in
* a 64-bit word if we want to enable a packed representation.
*
* There are two things that are special about the information here:
* - It's quicker to access. You have one fewer pointer hop, since finding the
* edata_t associated with an item always requires accessing the rtree leaf in
* which this data is stored.
* - It can be read unsynchronized, and without worrying about lifetime issues.
*/
typedef struct edata_map_info_s edata_map_info_t;
struct edata_map_info_s {
bool slab;
szind_t szind;
};
typedef struct edata_cmp_summary_s edata_cmp_summary_t;
struct edata_cmp_summary_s {
uint64_t sn;
uintptr_t addr;
};
/* Extent (span of pages). Use accessor functions for e_* fields. */
typedef struct edata_s edata_t;
ph_structs(edata_avail, edata_t);
ph_structs(edata_heap, edata_t);
struct edata_s {
/*
* Bitfield containing several fields:
*
* a: arena_ind
* b: slab
* c: committed
* p: pai
* z: zeroed
* g: guarded
* t: state
* i: szind
* f: nfree
* s: bin_shard
*
* 00000000 ... 0000ssss ssffffff ffffiiii iiiitttg zpcbaaaa aaaaaaaa
*
* arena_ind: Arena from which this extent came, or all 1 bits if
* unassociated.
*
* slab: The slab flag indicates whether the extent is used for a slab
* of small regions. This helps differentiate small size classes,
* and it indicates whether interior pointers can be looked up via
* iealloc().
*
* committed: The committed flag indicates whether physical memory is
* committed to the extent, whether explicitly or implicitly
* as on a system that overcommits and satisfies physical
* memory needs on demand via soft page faults.
*
* pai: The pai flag is an extent_pai_t.
*
* zeroed: The zeroed flag is used by extent recycling code to track
* whether memory is zero-filled.
*
* guarded: The guarded flag is use by the sanitizer to track whether
* the extent has page guards around it.
*
* state: The state flag is an extent_state_t.
*
* szind: The szind flag indicates usable size class index for
* allocations residing in this extent, regardless of whether the
* extent is a slab. Extent size and usable size often differ
* even for non-slabs, either due to sz_large_pad or promotion of
* sampled small regions.
*
* nfree: Number of free regions in slab.
*
* bin_shard: the shard of the bin from which this extent came.
*/
uint64_t e_bits;
#define MASK(CURRENT_FIELD_WIDTH, CURRENT_FIELD_SHIFT) ((((((uint64_t)0x1U) << (CURRENT_FIELD_WIDTH)) - 1)) << (CURRENT_FIELD_SHIFT))
#define EDATA_BITS_ARENA_WIDTH MALLOCX_ARENA_BITS
#define EDATA_BITS_ARENA_SHIFT 0
#define EDATA_BITS_ARENA_MASK MASK(EDATA_BITS_ARENA_WIDTH, EDATA_BITS_ARENA_SHIFT)
#define EDATA_BITS_SLAB_WIDTH 1
#define EDATA_BITS_SLAB_SHIFT (EDATA_BITS_ARENA_WIDTH + EDATA_BITS_ARENA_SHIFT)
#define EDATA_BITS_SLAB_MASK MASK(EDATA_BITS_SLAB_WIDTH, EDATA_BITS_SLAB_SHIFT)
#define EDATA_BITS_COMMITTED_WIDTH 1
#define EDATA_BITS_COMMITTED_SHIFT (EDATA_BITS_SLAB_WIDTH + EDATA_BITS_SLAB_SHIFT)
#define EDATA_BITS_COMMITTED_MASK MASK(EDATA_BITS_COMMITTED_WIDTH, EDATA_BITS_COMMITTED_SHIFT)
#define EDATA_BITS_PAI_WIDTH 1
#define EDATA_BITS_PAI_SHIFT (EDATA_BITS_COMMITTED_WIDTH + EDATA_BITS_COMMITTED_SHIFT)
#define EDATA_BITS_PAI_MASK MASK(EDATA_BITS_PAI_WIDTH, EDATA_BITS_PAI_SHIFT)
#define EDATA_BITS_ZEROED_WIDTH 1
#define EDATA_BITS_ZEROED_SHIFT (EDATA_BITS_PAI_WIDTH + EDATA_BITS_PAI_SHIFT)
#define EDATA_BITS_ZEROED_MASK MASK(EDATA_BITS_ZEROED_WIDTH, EDATA_BITS_ZEROED_SHIFT)
#define EDATA_BITS_GUARDED_WIDTH 1
#define EDATA_BITS_GUARDED_SHIFT (EDATA_BITS_ZEROED_WIDTH + EDATA_BITS_ZEROED_SHIFT)
#define EDATA_BITS_GUARDED_MASK MASK(EDATA_BITS_GUARDED_WIDTH, EDATA_BITS_GUARDED_SHIFT)
#define EDATA_BITS_STATE_WIDTH 3
#define EDATA_BITS_STATE_SHIFT (EDATA_BITS_GUARDED_WIDTH + EDATA_BITS_GUARDED_SHIFT)
#define EDATA_BITS_STATE_MASK MASK(EDATA_BITS_STATE_WIDTH, EDATA_BITS_STATE_SHIFT)
#define EDATA_BITS_SZIND_WIDTH LG_CEIL(SC_NSIZES)
#define EDATA_BITS_SZIND_SHIFT (EDATA_BITS_STATE_WIDTH + EDATA_BITS_STATE_SHIFT)
#define EDATA_BITS_SZIND_MASK MASK(EDATA_BITS_SZIND_WIDTH, EDATA_BITS_SZIND_SHIFT)
#define EDATA_BITS_NFREE_WIDTH (SC_LG_SLAB_MAXREGS + 1)
#define EDATA_BITS_NFREE_SHIFT (EDATA_BITS_SZIND_WIDTH + EDATA_BITS_SZIND_SHIFT)
#define EDATA_BITS_NFREE_MASK MASK(EDATA_BITS_NFREE_WIDTH, EDATA_BITS_NFREE_SHIFT)
#define EDATA_BITS_BINSHARD_WIDTH 6
#define EDATA_BITS_BINSHARD_SHIFT (EDATA_BITS_NFREE_WIDTH + EDATA_BITS_NFREE_SHIFT)
#define EDATA_BITS_BINSHARD_MASK MASK(EDATA_BITS_BINSHARD_WIDTH, EDATA_BITS_BINSHARD_SHIFT)
#define EDATA_BITS_IS_HEAD_WIDTH 1
#define EDATA_BITS_IS_HEAD_SHIFT (EDATA_BITS_BINSHARD_WIDTH + EDATA_BITS_BINSHARD_SHIFT)
#define EDATA_BITS_IS_HEAD_MASK MASK(EDATA_BITS_IS_HEAD_WIDTH, EDATA_BITS_IS_HEAD_SHIFT)
/* Pointer to the extent that this structure is responsible for. */
void *e_addr;
union {
/*
* Extent size and serial number associated with the extent
* structure (different than the serial number for the extent at
* e_addr).
*
* ssssssss [...] ssssssss ssssnnnn nnnnnnnn
*/
size_t e_size_esn;
#define EDATA_SIZE_MASK ((size_t)~(PAGE-1))
#define EDATA_ESN_MASK ((size_t)PAGE-1)
/* Base extent size, which may not be a multiple of PAGE. */
size_t e_bsize;
};
/*
* If this edata is a user allocation from an HPA, it comes out of some
* pageslab (we don't yet support huegpage allocations that don't fit
* into pageslabs). This tracks it.
*/
hpdata_t *e_ps;
/*
* Serial number. These are not necessarily unique; splitting an extent
* results in two extents with the same serial number.
*/
uint64_t e_sn;
union {
/*
* List linkage used when the edata_t is active; either in
* arena's large allocations or bin_t's slabs_full.
*/
ql_elm(edata_t) ql_link_active;
/*
* Pairing heap linkage. Used whenever the extent is inactive
* (in the page allocators), or when it is active and in
* slabs_nonfull, or when the edata_t is unassociated with an
* extent and sitting in an edata_cache.
*/
union {
edata_heap_link_t heap_link;
edata_avail_link_t avail_link;
};
};
union {
/*
* List linkage used when the extent is inactive:
* - Stashed dirty extents
* - Ecache LRU functionality.
*/
ql_elm(edata_t) ql_link_inactive;
/* Small region slab metadata. */
slab_data_t e_slab_data;
/* Profiling data, used for large objects. */
e_prof_info_t e_prof_info;
};
};
TYPED_LIST(edata_list_active, edata_t, ql_link_active)
TYPED_LIST(edata_list_inactive, edata_t, ql_link_inactive)
static inline unsigned
edata_arena_ind_get(const edata_t *edata) {
unsigned arena_ind = (unsigned)((edata->e_bits &
EDATA_BITS_ARENA_MASK) >> EDATA_BITS_ARENA_SHIFT);
assert(arena_ind < MALLOCX_ARENA_LIMIT);
return arena_ind;
}
static inline szind_t
edata_szind_get_maybe_invalid(const edata_t *edata) {
szind_t szind = (szind_t)((edata->e_bits & EDATA_BITS_SZIND_MASK) >>
EDATA_BITS_SZIND_SHIFT);
assert(szind <= SC_NSIZES);
return szind;
}
static inline szind_t
edata_szind_get(const edata_t *edata) {
szind_t szind = edata_szind_get_maybe_invalid(edata);
assert(szind < SC_NSIZES); /* Never call when "invalid". */
return szind;
}
static inline size_t
edata_usize_get(const edata_t *edata) {
return sz_index2size(edata_szind_get(edata));
}
static inline unsigned
edata_binshard_get(const edata_t *edata) {
unsigned binshard = (unsigned)((edata->e_bits &
EDATA_BITS_BINSHARD_MASK) >> EDATA_BITS_BINSHARD_SHIFT);
assert(binshard < bin_infos[edata_szind_get(edata)].n_shards);
return binshard;
}
static inline uint64_t
edata_sn_get(const edata_t *edata) {
return edata->e_sn;
}
static inline extent_state_t
edata_state_get(const edata_t *edata) {
return (extent_state_t)((edata->e_bits & EDATA_BITS_STATE_MASK) >>
EDATA_BITS_STATE_SHIFT);
}
static inline bool
edata_guarded_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_GUARDED_MASK) >>
EDATA_BITS_GUARDED_SHIFT);
}
static inline bool
edata_zeroed_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_ZEROED_MASK) >>
EDATA_BITS_ZEROED_SHIFT);
}
static inline bool
edata_committed_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_COMMITTED_MASK) >>
EDATA_BITS_COMMITTED_SHIFT);
}
static inline extent_pai_t
edata_pai_get(const edata_t *edata) {
return (extent_pai_t)((edata->e_bits & EDATA_BITS_PAI_MASK) >>
EDATA_BITS_PAI_SHIFT);
}
static inline bool
edata_slab_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_SLAB_MASK) >>
EDATA_BITS_SLAB_SHIFT);
}
static inline unsigned
edata_nfree_get(const edata_t *edata) {
assert(edata_slab_get(edata));
return (unsigned)((edata->e_bits & EDATA_BITS_NFREE_MASK) >>
EDATA_BITS_NFREE_SHIFT);
}
static inline void *
edata_base_get(const edata_t *edata) {
assert(edata->e_addr == PAGE_ADDR2BASE(edata->e_addr) ||
!edata_slab_get(edata));
return PAGE_ADDR2BASE(edata->e_addr);
}
static inline void *
edata_addr_get(const edata_t *edata) {
assert(edata->e_addr == PAGE_ADDR2BASE(edata->e_addr) ||
!edata_slab_get(edata));
return edata->e_addr;
}
static inline size_t
edata_size_get(const edata_t *edata) {
return (edata->e_size_esn & EDATA_SIZE_MASK);
}
static inline size_t
edata_esn_get(const edata_t *edata) {
return (edata->e_size_esn & EDATA_ESN_MASK);
}
static inline size_t
edata_bsize_get(const edata_t *edata) {
return edata->e_bsize;
}
static inline hpdata_t *
edata_ps_get(const edata_t *edata) {
assert(edata_pai_get(edata) == EXTENT_PAI_HPA);
return edata->e_ps;
}
static inline void *
edata_before_get(const edata_t *edata) {
return (void *)((uintptr_t)edata_base_get(edata) - PAGE);
}
static inline void *
edata_last_get(const edata_t *edata) {
return (void *)((uintptr_t)edata_base_get(edata) +
edata_size_get(edata) - PAGE);
}
static inline void *
edata_past_get(const edata_t *edata) {
return (void *)((uintptr_t)edata_base_get(edata) +
edata_size_get(edata));
}
static inline slab_data_t *
edata_slab_data_get(edata_t *edata) {
assert(edata_slab_get(edata));
return &edata->e_slab_data;
}
static inline const slab_data_t *
edata_slab_data_get_const(const edata_t *edata) {
assert(edata_slab_get(edata));
return &edata->e_slab_data;
}
static inline prof_tctx_t *
edata_prof_tctx_get(const edata_t *edata) {
return (prof_tctx_t *)atomic_load_p(&edata->e_prof_info.e_prof_tctx,
ATOMIC_ACQUIRE);
}
static inline const nstime_t *
edata_prof_alloc_time_get(const edata_t *edata) {
return &edata->e_prof_info.e_prof_alloc_time;
}
static inline size_t
edata_prof_alloc_size_get(const edata_t *edata) {
return edata->e_prof_info.e_prof_alloc_size;
}
static inline prof_recent_t *
edata_prof_recent_alloc_get_dont_call_directly(const edata_t *edata) {
return (prof_recent_t *)atomic_load_p(
&edata->e_prof_info.e_prof_recent_alloc, ATOMIC_RELAXED);
}
static inline void
edata_arena_ind_set(edata_t *edata, unsigned arena_ind) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_ARENA_MASK) |
((uint64_t)arena_ind << EDATA_BITS_ARENA_SHIFT);
}
static inline void
edata_binshard_set(edata_t *edata, unsigned binshard) {
/* The assertion assumes szind is set already. */
assert(binshard < bin_infos[edata_szind_get(edata)].n_shards);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_BINSHARD_MASK) |
((uint64_t)binshard << EDATA_BITS_BINSHARD_SHIFT);
}
static inline void
edata_addr_set(edata_t *edata, void *addr) {
edata->e_addr = addr;
}
static inline void
edata_size_set(edata_t *edata, size_t size) {
assert((size & ~EDATA_SIZE_MASK) == 0);
edata->e_size_esn = size | (edata->e_size_esn & ~EDATA_SIZE_MASK);
}
static inline void
edata_esn_set(edata_t *edata, size_t esn) {
edata->e_size_esn = (edata->e_size_esn & ~EDATA_ESN_MASK) | (esn &
EDATA_ESN_MASK);
}
static inline void
edata_bsize_set(edata_t *edata, size_t bsize) {
edata->e_bsize = bsize;
}
static inline void
edata_ps_set(edata_t *edata, hpdata_t *ps) {
assert(edata_pai_get(edata) == EXTENT_PAI_HPA);
edata->e_ps = ps;
}
static inline void
edata_szind_set(edata_t *edata, szind_t szind) {
assert(szind <= SC_NSIZES); /* SC_NSIZES means "invalid". */
edata->e_bits = (edata->e_bits & ~EDATA_BITS_SZIND_MASK) |
((uint64_t)szind << EDATA_BITS_SZIND_SHIFT);
}
static inline void
edata_nfree_set(edata_t *edata, unsigned nfree) {
assert(edata_slab_get(edata));
edata->e_bits = (edata->e_bits & ~EDATA_BITS_NFREE_MASK) |
((uint64_t)nfree << EDATA_BITS_NFREE_SHIFT);
}
static inline void
edata_nfree_binshard_set(edata_t *edata, unsigned nfree, unsigned binshard) {
/* The assertion assumes szind is set already. */
assert(binshard < bin_infos[edata_szind_get(edata)].n_shards);
edata->e_bits = (edata->e_bits &
(~EDATA_BITS_NFREE_MASK & ~EDATA_BITS_BINSHARD_MASK)) |
((uint64_t)binshard << EDATA_BITS_BINSHARD_SHIFT) |
((uint64_t)nfree << EDATA_BITS_NFREE_SHIFT);
}
static inline void
edata_nfree_inc(edata_t *edata) {
assert(edata_slab_get(edata));
edata->e_bits += ((uint64_t)1U << EDATA_BITS_NFREE_SHIFT);
}
static inline void
edata_nfree_dec(edata_t *edata) {
assert(edata_slab_get(edata));
edata->e_bits -= ((uint64_t)1U << EDATA_BITS_NFREE_SHIFT);
}
static inline void
edata_nfree_sub(edata_t *edata, uint64_t n) {
assert(edata_slab_get(edata));
edata->e_bits -= (n << EDATA_BITS_NFREE_SHIFT);
}
static inline void
edata_sn_set(edata_t *edata, uint64_t sn) {
edata->e_sn = sn;
}
static inline void
edata_state_set(edata_t *edata, extent_state_t state) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_STATE_MASK) |
((uint64_t)state << EDATA_BITS_STATE_SHIFT);
}
static inline void
edata_guarded_set(edata_t *edata, bool guarded) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_GUARDED_MASK) |
((uint64_t)guarded << EDATA_BITS_GUARDED_SHIFT);
}
static inline void
edata_zeroed_set(edata_t *edata, bool zeroed) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_ZEROED_MASK) |
((uint64_t)zeroed << EDATA_BITS_ZEROED_SHIFT);
}
static inline void
edata_committed_set(edata_t *edata, bool committed) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_COMMITTED_MASK) |
((uint64_t)committed << EDATA_BITS_COMMITTED_SHIFT);
}
static inline void
edata_pai_set(edata_t *edata, extent_pai_t pai) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_PAI_MASK) |
((uint64_t)pai << EDATA_BITS_PAI_SHIFT);
}
static inline void
edata_slab_set(edata_t *edata, bool slab) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_SLAB_MASK) |
((uint64_t)slab << EDATA_BITS_SLAB_SHIFT);
}
static inline void
edata_prof_tctx_set(edata_t *edata, prof_tctx_t *tctx) {
atomic_store_p(&edata->e_prof_info.e_prof_tctx, tctx, ATOMIC_RELEASE);
}
static inline void
edata_prof_alloc_time_set(edata_t *edata, nstime_t *t) {
nstime_copy(&edata->e_prof_info.e_prof_alloc_time, t);
}
static inline void
edata_prof_alloc_size_set(edata_t *edata, size_t size) {
edata->e_prof_info.e_prof_alloc_size = size;
}
static inline void
edata_prof_recent_alloc_set_dont_call_directly(edata_t *edata,
prof_recent_t *recent_alloc) {
atomic_store_p(&edata->e_prof_info.e_prof_recent_alloc, recent_alloc,
ATOMIC_RELAXED);
}
static inline bool
edata_is_head_get(edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_IS_HEAD_MASK) >>
EDATA_BITS_IS_HEAD_SHIFT);
}
static inline void
edata_is_head_set(edata_t *edata, bool is_head) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_IS_HEAD_MASK) |
((uint64_t)is_head << EDATA_BITS_IS_HEAD_SHIFT);
}
static inline bool
edata_state_in_transition(extent_state_t state) {
return state >= extent_state_transition;
}
/*
* Because this function is implemented as a sequence of bitfield modifications,
* even though each individual bit is properly initialized, we technically read
* uninitialized data within it. This is mostly fine, since most callers get
* their edatas from zeroing sources, but callers who make stack edata_ts need
* to manually zero them.
*/
static inline void
edata_init(edata_t *edata, unsigned arena_ind, void *addr, size_t size,
bool slab, szind_t szind, uint64_t sn, extent_state_t state, bool zeroed,
bool committed, extent_pai_t pai, extent_head_state_t is_head) {
assert(addr == PAGE_ADDR2BASE(addr) || !slab);
edata_arena_ind_set(edata, arena_ind);
edata_addr_set(edata, addr);
edata_size_set(edata, size);
edata_slab_set(edata, slab);
edata_szind_set(edata, szind);
edata_sn_set(edata, sn);
edata_state_set(edata, state);
edata_guarded_set(edata, false);
edata_zeroed_set(edata, zeroed);
edata_committed_set(edata, committed);
edata_pai_set(edata, pai);
edata_is_head_set(edata, is_head == EXTENT_IS_HEAD);
if (config_prof) {
edata_prof_tctx_set(edata, NULL);
}
}
static inline void
edata_binit(edata_t *edata, void *addr, size_t bsize, uint64_t sn) {
edata_arena_ind_set(edata, (1U << MALLOCX_ARENA_BITS) - 1);
edata_addr_set(edata, addr);
edata_bsize_set(edata, bsize);
edata_slab_set(edata, false);
edata_szind_set(edata, SC_NSIZES);
edata_sn_set(edata, sn);
edata_state_set(edata, extent_state_active);
edata_guarded_set(edata, false);
edata_zeroed_set(edata, true);
edata_committed_set(edata, true);
/*
* This isn't strictly true, but base allocated extents never get
* deallocated and can't be looked up in the emap, but no sense in
* wasting a state bit to encode this fact.
*/
edata_pai_set(edata, EXTENT_PAI_PAC);
}
static inline int
edata_esn_comp(const edata_t *a, const edata_t *b) {
size_t a_esn = edata_esn_get(a);
size_t b_esn = edata_esn_get(b);
return (a_esn > b_esn) - (a_esn < b_esn);
}
static inline int
edata_ead_comp(const edata_t *a, const edata_t *b) {
uintptr_t a_eaddr = (uintptr_t)a;
uintptr_t b_eaddr = (uintptr_t)b;
return (a_eaddr > b_eaddr) - (a_eaddr < b_eaddr);
}
static inline edata_cmp_summary_t
edata_cmp_summary_get(const edata_t *edata) {
return (edata_cmp_summary_t){edata_sn_get(edata),
(uintptr_t)edata_addr_get(edata)};
}
static inline int
edata_cmp_summary_comp(edata_cmp_summary_t a, edata_cmp_summary_t b) {
int ret;
ret = (a.sn > b.sn) - (a.sn < b.sn);
if (ret != 0) {
return ret;
}
ret = (a.addr > b.addr) - (a.addr < b.addr);
return ret;
}
static inline int
edata_snad_comp(const edata_t *a, const edata_t *b) {
edata_cmp_summary_t a_cmp = edata_cmp_summary_get(a);
edata_cmp_summary_t b_cmp = edata_cmp_summary_get(b);
return edata_cmp_summary_comp(a_cmp, b_cmp);
}
static inline int
edata_esnead_comp(const edata_t *a, const edata_t *b) {
int ret;
ret = edata_esn_comp(a, b);
if (ret != 0) {
return ret;
}
ret = edata_ead_comp(a, b);
return ret;
}
ph_proto(, edata_avail, edata_t)
ph_proto(, edata_heap, edata_t)
#endif /* JEMALLOC_INTERNAL_EDATA_H */
#ifndef JEMALLOC_INTERNAL_EDATA_CACHE_H
#define JEMALLOC_INTERNAL_EDATA_CACHE_H
#include "jemalloc/internal/base.h"
/* For tests only. */
#define EDATA_CACHE_FAST_FILL 4
/*
* A cache of edata_t structures allocated via base_alloc_edata (as opposed to
* the underlying extents they describe). The contents of returned edata_t
* objects are garbage and cannot be relied upon.
*/
typedef struct edata_cache_s edata_cache_t;
struct edata_cache_s {
edata_avail_t avail;
atomic_zu_t count;
malloc_mutex_t mtx;
base_t *base;
};
bool edata_cache_init(edata_cache_t *edata_cache, base_t *base);
edata_t *edata_cache_get(tsdn_t *tsdn, edata_cache_t *edata_cache);
void edata_cache_put(tsdn_t *tsdn, edata_cache_t *edata_cache, edata_t *edata);
void edata_cache_prefork(tsdn_t *tsdn, edata_cache_t *edata_cache);
void edata_cache_postfork_parent(tsdn_t *tsdn, edata_cache_t *edata_cache);
void edata_cache_postfork_child(tsdn_t *tsdn, edata_cache_t *edata_cache);
/*
* An edata_cache_small is like an edata_cache, but it relies on external
* synchronization and avoids first-fit strategies.
*/
typedef struct edata_cache_fast_s edata_cache_fast_t;
struct edata_cache_fast_s {
edata_list_inactive_t list;
edata_cache_t *fallback;
bool disabled;
};
void edata_cache_fast_init(edata_cache_fast_t *ecs, edata_cache_t *fallback);
edata_t *edata_cache_fast_get(tsdn_t *tsdn, edata_cache_fast_t *ecs);
void edata_cache_fast_put(tsdn_t *tsdn, edata_cache_fast_t *ecs,
edata_t *edata);
void edata_cache_fast_disable(tsdn_t *tsdn, edata_cache_fast_t *ecs);
#endif /* JEMALLOC_INTERNAL_EDATA_CACHE_H */
#ifndef JEMALLOC_INTERNAL_EHOOKS_H
#define JEMALLOC_INTERNAL_EHOOKS_H
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/extent_mmap.h"
/*
* This module is the internal interface to the extent hooks (both
* user-specified and external). Eventually, this will give us the flexibility
* to use multiple different versions of user-visible extent-hook APIs under a
* single user interface.
*
* Current API expansions (not available to anyone but the default hooks yet):
* - Head state tracking. Hooks can decide whether or not to merge two
* extents based on whether or not one of them is the head (i.e. was
* allocated on its own). The later extent loses its "head" status.
*/
extern const extent_hooks_t ehooks_default_extent_hooks;
typedef struct ehooks_s ehooks_t;
struct ehooks_s {
/*
* The user-visible id that goes with the ehooks (i.e. that of the base
* they're a part of, the associated arena's index within the arenas
* array).
*/
unsigned ind;
/* Logically an extent_hooks_t *. */
atomic_p_t ptr;
};
extern const extent_hooks_t ehooks_default_extent_hooks;
/*
* These are not really part of the public API. Each hook has a fast-path for
* the default-hooks case that can avoid various small inefficiencies:
* - Forgetting tsd and then calling tsd_get within the hook.
* - Getting more state than necessary out of the extent_t.
* - Doing arena_ind -> arena -> arena_ind lookups.
* By making the calls to these functions visible to the compiler, it can move
* those extra bits of computation down below the fast-paths where they get ignored.
*/
void *ehooks_default_alloc_impl(tsdn_t *tsdn, void *new_addr, size_t size,
size_t alignment, bool *zero, bool *commit, unsigned arena_ind);
bool ehooks_default_dalloc_impl(void *addr, size_t size);
void ehooks_default_destroy_impl(void *addr, size_t size);
bool ehooks_default_commit_impl(void *addr, size_t offset, size_t length);
bool ehooks_default_decommit_impl(void *addr, size_t offset, size_t length);
#ifdef PAGES_CAN_PURGE_LAZY
bool ehooks_default_purge_lazy_impl(void *addr, size_t offset, size_t length);
#endif
#ifdef PAGES_CAN_PURGE_FORCED
bool ehooks_default_purge_forced_impl(void *addr, size_t offset, size_t length);
#endif
bool ehooks_default_split_impl();
/*
* Merge is the only default extent hook we declare -- see the comment in
* ehooks_merge.
*/
bool ehooks_default_merge(extent_hooks_t *extent_hooks, void *addr_a,
size_t size_a, void *addr_b, size_t size_b, bool committed,
unsigned arena_ind);
bool ehooks_default_merge_impl(tsdn_t *tsdn, void *addr_a, void *addr_b);
void ehooks_default_zero_impl(void *addr, size_t size);
void ehooks_default_guard_impl(void *guard1, void *guard2);
void ehooks_default_unguard_impl(void *guard1, void *guard2);
/*
* We don't officially support reentrancy from wtihin the extent hooks. But
* various people who sit within throwing distance of the jemalloc team want
* that functionality in certain limited cases. The default reentrancy guards
* assert that we're not reentrant from a0 (since it's the bootstrap arena,
* where reentrant allocations would be redirected), which we would incorrectly
* trigger in cases where a0 has extent hooks (those hooks themselves can't be
* reentrant, then, but there are reasonable uses for such functionality, like
* putting internal metadata on hugepages). Therefore, we use the raw
* reentrancy guards.
*
* Eventually, we need to think more carefully about whether and where we
* support allocating from within extent hooks (and what that means for things
* like profiling, stats collection, etc.), and document what the guarantee is.
*/
static inline void
ehooks_pre_reentrancy(tsdn_t *tsdn) {
tsd_t *tsd = tsdn_null(tsdn) ? tsd_fetch() : tsdn_tsd(tsdn);
tsd_pre_reentrancy_raw(tsd);
}
static inline void
ehooks_post_reentrancy(tsdn_t *tsdn) {
tsd_t *tsd = tsdn_null(tsdn) ? tsd_fetch() : tsdn_tsd(tsdn);
tsd_post_reentrancy_raw(tsd);
}
/* Beginning of the public API. */
void ehooks_init(ehooks_t *ehooks, extent_hooks_t *extent_hooks, unsigned ind);
static inline unsigned
ehooks_ind_get(const ehooks_t *ehooks) {
return ehooks->ind;
}
static inline void
ehooks_set_extent_hooks_ptr(ehooks_t *ehooks, extent_hooks_t *extent_hooks) {
atomic_store_p(&ehooks->ptr, extent_hooks, ATOMIC_RELEASE);
}
static inline extent_hooks_t *
ehooks_get_extent_hooks_ptr(ehooks_t *ehooks) {
return (extent_hooks_t *)atomic_load_p(&ehooks->ptr, ATOMIC_ACQUIRE);
}
static inline bool
ehooks_are_default(ehooks_t *ehooks) {
return ehooks_get_extent_hooks_ptr(ehooks) ==
&ehooks_default_extent_hooks;
}
/*
* In some cases, a caller needs to allocate resources before attempting to call
* a hook. If that hook is doomed to fail, this is wasteful. We therefore
* include some checks for such cases.
*/
static inline bool
ehooks_dalloc_will_fail(ehooks_t *ehooks) {
if (ehooks_are_default(ehooks)) {
return opt_retain;
} else {
return ehooks_get_extent_hooks_ptr(ehooks)->dalloc == NULL;
}
}
static inline bool
ehooks_split_will_fail(ehooks_t *ehooks) {
return ehooks_get_extent_hooks_ptr(ehooks)->split == NULL;
}
static inline bool
ehooks_merge_will_fail(ehooks_t *ehooks) {
return ehooks_get_extent_hooks_ptr(ehooks)->merge == NULL;
}
static inline bool
ehooks_guard_will_fail(ehooks_t *ehooks) {
/*
* Before the guard hooks are officially introduced, limit the use to
* the default hooks only.
*/
return !ehooks_are_default(ehooks);
}
/*
* Some hooks are required to return zeroed memory in certain situations. In
* debug mode, we do some heuristic checks that they did what they were supposed
* to.
*
* This isn't really ehooks-specific (i.e. anyone can check for zeroed memory).
* But incorrect zero information indicates an ehook bug.
*/
static inline void
ehooks_debug_zero_check(void *addr, size_t size) {
assert(((uintptr_t)addr & PAGE_MASK) == 0);
assert((size & PAGE_MASK) == 0);
assert(size > 0);
if (config_debug) {
/* Check the whole first page. */
size_t *p = (size_t *)addr;
for (size_t i = 0; i < PAGE / sizeof(size_t); i++) {
assert(p[i] == 0);
}
/*
* And 4 spots within. There's a tradeoff here; the larger
* this number, the more likely it is that we'll catch a bug
* where ehooks return a sparsely non-zero range. But
* increasing the number of checks also increases the number of
* page faults in debug mode. FreeBSD does much of their
* day-to-day development work in debug mode, so we don't want
* even the debug builds to be too slow.
*/
const size_t nchecks = 4;
assert(PAGE >= sizeof(size_t) * nchecks);
for (size_t i = 0; i < nchecks; ++i) {
assert(p[i * (size / sizeof(size_t) / nchecks)] == 0);
}
}
}
static inline void *
ehooks_alloc(tsdn_t *tsdn, ehooks_t *ehooks, void *new_addr, size_t size,
size_t alignment, bool *zero, bool *commit) {
bool orig_zero = *zero;
void *ret;
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
ret = ehooks_default_alloc_impl(tsdn, new_addr, size,
alignment, zero, commit, ehooks_ind_get(ehooks));
} else {
ehooks_pre_reentrancy(tsdn);
ret = extent_hooks->alloc(extent_hooks, new_addr, size,
alignment, zero, commit, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
}
assert(new_addr == NULL || ret == NULL || new_addr == ret);
assert(!orig_zero || *zero);
if (*zero && ret != NULL) {
ehooks_debug_zero_check(ret, size);
}
return ret;
}
static inline bool
ehooks_dalloc(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
bool committed) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
return ehooks_default_dalloc_impl(addr, size);
} else if (extent_hooks->dalloc == NULL) {
return true;
} else {
ehooks_pre_reentrancy(tsdn);
bool err = extent_hooks->dalloc(extent_hooks, addr, size,
committed, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
return err;
}
}
static inline void
ehooks_destroy(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
bool committed) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
ehooks_default_destroy_impl(addr, size);
} else if (extent_hooks->destroy == NULL) {
/* Do nothing. */
} else {
ehooks_pre_reentrancy(tsdn);
extent_hooks->destroy(extent_hooks, addr, size, committed,
ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
}
}
static inline bool
ehooks_commit(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
size_t offset, size_t length) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
bool err;
if (extent_hooks == &ehooks_default_extent_hooks) {
err = ehooks_default_commit_impl(addr, offset, length);
} else if (extent_hooks->commit == NULL) {
err = true;
} else {
ehooks_pre_reentrancy(tsdn);
err = extent_hooks->commit(extent_hooks, addr, size,
offset, length, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
}
if (!err) {
ehooks_debug_zero_check(addr, size);
}
return err;
}
static inline bool
ehooks_decommit(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
size_t offset, size_t length) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
return ehooks_default_decommit_impl(addr, offset, length);
} else if (extent_hooks->decommit == NULL) {
return true;
} else {
ehooks_pre_reentrancy(tsdn);
bool err = extent_hooks->decommit(extent_hooks, addr, size,
offset, length, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
return err;
}
}
static inline bool
ehooks_purge_lazy(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
size_t offset, size_t length) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
#ifdef PAGES_CAN_PURGE_LAZY
if (extent_hooks == &ehooks_default_extent_hooks) {
return ehooks_default_purge_lazy_impl(addr, offset, length);
}
#endif
if (extent_hooks->purge_lazy == NULL) {
return true;
} else {
ehooks_pre_reentrancy(tsdn);
bool err = extent_hooks->purge_lazy(extent_hooks, addr, size,
offset, length, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
return err;
}
}
static inline bool
ehooks_purge_forced(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
size_t offset, size_t length) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
/*
* It would be correct to have a ehooks_debug_zero_check call at the end
* of this function; purge_forced is required to zero. But checking
* would touch the page in question, which may have performance
* consequences (imagine the hooks are using hugepages, with a global
* zero page off). Even in debug mode, it's usually a good idea to
* avoid cases that can dramatically increase memory consumption.
*/
#ifdef PAGES_CAN_PURGE_FORCED
if (extent_hooks == &ehooks_default_extent_hooks) {
return ehooks_default_purge_forced_impl(addr, offset, length);
}
#endif
if (extent_hooks->purge_forced == NULL) {
return true;
} else {
ehooks_pre_reentrancy(tsdn);
bool err = extent_hooks->purge_forced(extent_hooks, addr, size,
offset, length, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
return err;
}
}
static inline bool
ehooks_split(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
size_t size_a, size_t size_b, bool committed) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (ehooks_are_default(ehooks)) {
return ehooks_default_split_impl();
} else if (extent_hooks->split == NULL) {
return true;
} else {
ehooks_pre_reentrancy(tsdn);
bool err = extent_hooks->split(extent_hooks, addr, size, size_a,
size_b, committed, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
return err;
}
}
static inline bool
ehooks_merge(tsdn_t *tsdn, ehooks_t *ehooks, void *addr_a, size_t size_a,
void *addr_b, size_t size_b, bool committed) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
return ehooks_default_merge_impl(tsdn, addr_a, addr_b);
} else if (extent_hooks->merge == NULL) {
return true;
} else {
ehooks_pre_reentrancy(tsdn);
bool err = extent_hooks->merge(extent_hooks, addr_a, size_a,
addr_b, size_b, committed, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
return err;
}
}
static inline void
ehooks_zero(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
ehooks_default_zero_impl(addr, size);
} else {
/*
* It would be correct to try using the user-provided purge
* hooks (since they are required to have zeroed the extent if
* they indicate success), but we don't necessarily know their
* cost. We'll be conservative and use memset.
*/
memset(addr, 0, size);
}
}
static inline bool
ehooks_guard(tsdn_t *tsdn, ehooks_t *ehooks, void *guard1, void *guard2) {
bool err;
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
ehooks_default_guard_impl(guard1, guard2);
err = false;
} else {
err = true;
}
return err;
}
static inline bool
ehooks_unguard(tsdn_t *tsdn, ehooks_t *ehooks, void *guard1, void *guard2) {
bool err;
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
ehooks_default_unguard_impl(guard1, guard2);
err = false;
} else {
err = true;
}
return err;
}
#endif /* JEMALLOC_INTERNAL_EHOOKS_H */
#ifndef JEMALLOC_INTERNAL_EMAP_H
#define JEMALLOC_INTERNAL_EMAP_H
#include "jemalloc/internal/base.h"
#include "jemalloc/internal/rtree.h"
/*
* Note: Ends without at semicolon, so that
* EMAP_DECLARE_RTREE_CTX;
* in uses will avoid empty-statement warnings.
*/
#define EMAP_DECLARE_RTREE_CTX \
rtree_ctx_t rtree_ctx_fallback; \
rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback)
typedef struct emap_s emap_t;
struct emap_s {
rtree_t rtree;
};
/* Used to pass rtree lookup context down the path. */
typedef struct emap_alloc_ctx_t emap_alloc_ctx_t;
struct emap_alloc_ctx_t {
szind_t szind;
bool slab;
};
typedef struct emap_full_alloc_ctx_s emap_full_alloc_ctx_t;
struct emap_full_alloc_ctx_s {
szind_t szind;
bool slab;
edata_t *edata;
};
bool emap_init(emap_t *emap, base_t *base, bool zeroed);
void emap_remap(tsdn_t *tsdn, emap_t *emap, edata_t *edata, szind_t szind,
bool slab);
void emap_update_edata_state(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
extent_state_t state);
/*
* The two acquire functions below allow accessing neighbor edatas, if it's safe
* and valid to do so (i.e. from the same arena, of the same state, etc.). This
* is necessary because the ecache locks are state based, and only protect
* edatas with the same state. Therefore the neighbor edata's state needs to be
* verified first, before chasing the edata pointer. The returned edata will be
* in an acquired state, meaning other threads will be prevented from accessing
* it, even if technically the edata can still be discovered from the rtree.
*
* This means, at any moment when holding pointers to edata, either one of the
* state based locks is held (and the edatas are all of the protected state), or
* the edatas are in an acquired state (e.g. in active or merging state). The
* acquire operation itself (changing the edata to an acquired state) is done
* under the state locks.
*/
edata_t *emap_try_acquire_edata_neighbor(tsdn_t *tsdn, emap_t *emap,
edata_t *edata, extent_pai_t pai, extent_state_t expected_state,
bool forward);
edata_t *emap_try_acquire_edata_neighbor_expand(tsdn_t *tsdn, emap_t *emap,
edata_t *edata, extent_pai_t pai, extent_state_t expected_state);
void emap_release_edata(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
extent_state_t new_state);
/*
* Associate the given edata with its beginning and end address, setting the
* szind and slab info appropriately.
* Returns true on error (i.e. resource exhaustion).
*/
bool emap_register_boundary(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
szind_t szind, bool slab);
/*
* Does the same thing, but with the interior of the range, for slab
* allocations.
*
* You might wonder why we don't just have a single emap_register function that
* does both depending on the value of 'slab'. The answer is twofold:
* - As a practical matter, in places like the extract->split->commit pathway,
* we defer the interior operation until we're sure that the commit won't fail
* (but we have to register the split boundaries there).
* - In general, we're trying to move to a world where the page-specific
* allocator doesn't know as much about how the pages it allocates will be
* used, and passing a 'slab' parameter everywhere makes that more
* complicated.
*
* Unlike the boundary version, this function can't fail; this is because slabs
* can't get big enough to touch a new page that neither of the boundaries
* touched, so no allocation is necessary to fill the interior once the boundary
* has been touched.
*/
void emap_register_interior(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
szind_t szind);
void emap_deregister_boundary(tsdn_t *tsdn, emap_t *emap, edata_t *edata);
void emap_deregister_interior(tsdn_t *tsdn, emap_t *emap, edata_t *edata);
typedef struct emap_prepare_s emap_prepare_t;
struct emap_prepare_s {
rtree_leaf_elm_t *lead_elm_a;
rtree_leaf_elm_t *lead_elm_b;
rtree_leaf_elm_t *trail_elm_a;
rtree_leaf_elm_t *trail_elm_b;
};
/**
* These functions the emap metadata management for merging, splitting, and
* reusing extents. In particular, they set the boundary mappings from
* addresses to edatas. If the result is going to be used as a slab, you
* still need to call emap_register_interior on it, though.
*
* Remap simply changes the szind and slab status of an extent's boundary
* mappings. If the extent is not a slab, it doesn't bother with updating the
* end mapping (since lookups only occur in the interior of an extent for
* slabs). Since the szind and slab status only make sense for active extents,
* this should only be called while activating or deactivating an extent.
*
* Split and merge have a "prepare" and a "commit" portion. The prepare portion
* does the operations that can be done without exclusive access to the extent
* in question, while the commit variant requires exclusive access to maintain
* the emap invariants. The only function that can fail is emap_split_prepare,
* and it returns true on failure (at which point the caller shouldn't commit).
*
* In all cases, "lead" refers to the lower-addressed extent, and trail to the
* higher-addressed one. It's the caller's responsibility to set the edata
* state appropriately.
*/
bool emap_split_prepare(tsdn_t *tsdn, emap_t *emap, emap_prepare_t *prepare,
edata_t *edata, size_t size_a, edata_t *trail, size_t size_b);
void emap_split_commit(tsdn_t *tsdn, emap_t *emap, emap_prepare_t *prepare,
edata_t *lead, size_t size_a, edata_t *trail, size_t size_b);
void emap_merge_prepare(tsdn_t *tsdn, emap_t *emap, emap_prepare_t *prepare,
edata_t *lead, edata_t *trail);
void emap_merge_commit(tsdn_t *tsdn, emap_t *emap, emap_prepare_t *prepare,
edata_t *lead, edata_t *trail);
/* Assert that the emap's view of the given edata matches the edata's view. */
void emap_do_assert_mapped(tsdn_t *tsdn, emap_t *emap, edata_t *edata);
static inline void
emap_assert_mapped(tsdn_t *tsdn, emap_t *emap, edata_t *edata) {
if (config_debug) {
emap_do_assert_mapped(tsdn, emap, edata);
}
}
/* Assert that the given edata isn't in the map. */
void emap_do_assert_not_mapped(tsdn_t *tsdn, emap_t *emap, edata_t *edata);
static inline void
emap_assert_not_mapped(tsdn_t *tsdn, emap_t *emap, edata_t *edata) {
if (config_debug) {
emap_do_assert_not_mapped(tsdn, emap, edata);
}
}
JEMALLOC_ALWAYS_INLINE bool
emap_edata_in_transition(tsdn_t *tsdn, emap_t *emap, edata_t *edata) {
assert(config_debug);
emap_assert_mapped(tsdn, emap, edata);
EMAP_DECLARE_RTREE_CTX;
rtree_contents_t contents = rtree_read(tsdn, &emap->rtree, rtree_ctx,
(uintptr_t)edata_base_get(edata));
return edata_state_in_transition(contents.metadata.state);
}
JEMALLOC_ALWAYS_INLINE bool
emap_edata_is_acquired(tsdn_t *tsdn, emap_t *emap, edata_t *edata) {
if (!config_debug) {
/* For assertions only. */
return false;
}
/*
* The edata is considered acquired if no other threads will attempt to
* read / write any fields from it. This includes a few cases:
*
* 1) edata not hooked into emap yet -- This implies the edata just got
* allocated or initialized.
*
* 2) in an active or transition state -- In both cases, the edata can
* be discovered from the emap, however the state tracked in the rtree
* will prevent other threads from accessing the actual edata.
*/
EMAP_DECLARE_RTREE_CTX;
rtree_leaf_elm_t *elm = rtree_leaf_elm_lookup(tsdn, &emap->rtree,
rtree_ctx, (uintptr_t)edata_base_get(edata), /* dependent */ true,
/* init_missing */ false);
if (elm == NULL) {
return true;
}
rtree_contents_t contents = rtree_leaf_elm_read(tsdn, &emap->rtree, elm,
/* dependent */ true);
if (contents.edata == NULL ||
contents.metadata.state == extent_state_active ||
edata_state_in_transition(contents.metadata.state)) {
return true;
}
return false;
}
JEMALLOC_ALWAYS_INLINE void
extent_assert_can_coalesce(const edata_t *inner, const edata_t *outer) {
assert(edata_arena_ind_get(inner) == edata_arena_ind_get(outer));
assert(edata_pai_get(inner) == edata_pai_get(outer));
assert(edata_committed_get(inner) == edata_committed_get(outer));
assert(edata_state_get(inner) == extent_state_active);
assert(edata_state_get(outer) == extent_state_merging);
assert(!edata_guarded_get(inner) && !edata_guarded_get(outer));
assert(edata_base_get(inner) == edata_past_get(outer) ||
edata_base_get(outer) == edata_past_get(inner));
}
JEMALLOC_ALWAYS_INLINE void
extent_assert_can_expand(const edata_t *original, const edata_t *expand) {
assert(edata_arena_ind_get(original) == edata_arena_ind_get(expand));
assert(edata_pai_get(original) == edata_pai_get(expand));
assert(edata_state_get(original) == extent_state_active);
assert(edata_state_get(expand) == extent_state_merging);
assert(edata_past_get(original) == edata_base_get(expand));
}
JEMALLOC_ALWAYS_INLINE edata_t *
emap_edata_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr) {
EMAP_DECLARE_RTREE_CTX;
return rtree_read(tsdn, &emap->rtree, rtree_ctx, (uintptr_t)ptr).edata;
}
/* Fills in alloc_ctx with the info in the map. */
JEMALLOC_ALWAYS_INLINE void
emap_alloc_ctx_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr,
emap_alloc_ctx_t *alloc_ctx) {
EMAP_DECLARE_RTREE_CTX;
rtree_metadata_t metadata = rtree_metadata_read(tsdn, &emap->rtree,
rtree_ctx, (uintptr_t)ptr);
alloc_ctx->szind = metadata.szind;
alloc_ctx->slab = metadata.slab;
}
/* The pointer must be mapped. */
JEMALLOC_ALWAYS_INLINE void
emap_full_alloc_ctx_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr,
emap_full_alloc_ctx_t *full_alloc_ctx) {
EMAP_DECLARE_RTREE_CTX;
rtree_contents_t contents = rtree_read(tsdn, &emap->rtree, rtree_ctx,
(uintptr_t)ptr);
full_alloc_ctx->edata = contents.edata;
full_alloc_ctx->szind = contents.metadata.szind;
full_alloc_ctx->slab = contents.metadata.slab;
}
/*
* The pointer is allowed to not be mapped.
*
* Returns true when the pointer is not present.
*/
JEMALLOC_ALWAYS_INLINE bool
emap_full_alloc_ctx_try_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr,
emap_full_alloc_ctx_t *full_alloc_ctx) {
EMAP_DECLARE_RTREE_CTX;
rtree_contents_t contents;
bool err = rtree_read_independent(tsdn, &emap->rtree, rtree_ctx,
(uintptr_t)ptr, &contents);
if (err) {
return true;
}
full_alloc_ctx->edata = contents.edata;
full_alloc_ctx->szind = contents.metadata.szind;
full_alloc_ctx->slab = contents.metadata.slab;
return false;
}
/*
* Only used on the fastpath of free. Returns true when cannot be fulfilled by
* fast path, e.g. when the metadata key is not cached.
*/
JEMALLOC_ALWAYS_INLINE bool
emap_alloc_ctx_try_lookup_fast(tsd_t *tsd, emap_t *emap, const void *ptr,
emap_alloc_ctx_t *alloc_ctx) {
/* Use the unsafe getter since this may gets called during exit. */
rtree_ctx_t *rtree_ctx = tsd_rtree_ctxp_get_unsafe(tsd);
rtree_metadata_t metadata;
bool err = rtree_metadata_try_read_fast(tsd_tsdn(tsd), &emap->rtree,
rtree_ctx, (uintptr_t)ptr, &metadata);
if (err) {
return true;
}
alloc_ctx->szind = metadata.szind;
alloc_ctx->slab = metadata.slab;
return false;
}
/*
* We want to do batch lookups out of the cache bins, which use
* cache_bin_ptr_array_get to access the i'th element of the bin (since they
* invert usual ordering in deciding what to flush). This lets the emap avoid
* caring about its caller's ordering.
*/
typedef const void *(*emap_ptr_getter)(void *ctx, size_t ind);
/*
* This allows size-checking assertions, which we can only do while we're in the
* process of edata lookups.
*/
typedef void (*emap_metadata_visitor)(void *ctx, emap_full_alloc_ctx_t *alloc_ctx);
typedef union emap_batch_lookup_result_u emap_batch_lookup_result_t;
union emap_batch_lookup_result_u {
edata_t *edata;
rtree_leaf_elm_t *rtree_leaf;
};
JEMALLOC_ALWAYS_INLINE void
emap_edata_lookup_batch(tsd_t *tsd, emap_t *emap, size_t nptrs,
emap_ptr_getter ptr_getter, void *ptr_getter_ctx,
emap_metadata_visitor metadata_visitor, void *metadata_visitor_ctx,
emap_batch_lookup_result_t *result) {
/* Avoids null-checking tsdn in the loop below. */
util_assume(tsd != NULL);
rtree_ctx_t *rtree_ctx = tsd_rtree_ctxp_get(tsd);
for (size_t i = 0; i < nptrs; i++) {
const void *ptr = ptr_getter(ptr_getter_ctx, i);
/*
* Reuse the edatas array as a temp buffer, lying a little about
* the types.
*/
result[i].rtree_leaf = rtree_leaf_elm_lookup(tsd_tsdn(tsd),
&emap->rtree, rtree_ctx, (uintptr_t)ptr,
/* dependent */ true, /* init_missing */ false);
}
for (size_t i = 0; i < nptrs; i++) {
rtree_leaf_elm_t *elm = result[i].rtree_leaf;
rtree_contents_t contents = rtree_leaf_elm_read(tsd_tsdn(tsd),
&emap->rtree, elm, /* dependent */ true);
result[i].edata = contents.edata;
emap_full_alloc_ctx_t alloc_ctx;
/*
* Not all these fields are read in practice by the metadata
* visitor. But the compiler can easily optimize away the ones
* that aren't, so no sense in being incomplete.
*/
alloc_ctx.szind = contents.metadata.szind;
alloc_ctx.slab = contents.metadata.slab;
alloc_ctx.edata = contents.edata;
metadata_visitor(metadata_visitor_ctx, &alloc_ctx);
}
}
#endif /* JEMALLOC_INTERNAL_EMAP_H */
...@@ -6,6 +6,7 @@ ...@@ -6,6 +6,7 @@
typedef enum emitter_output_e emitter_output_t; typedef enum emitter_output_e emitter_output_t;
enum emitter_output_e { enum emitter_output_e {
emitter_output_json, emitter_output_json,
emitter_output_json_compact,
emitter_output_table emitter_output_table
}; };
...@@ -21,6 +22,7 @@ typedef enum emitter_type_e emitter_type_t; ...@@ -21,6 +22,7 @@ typedef enum emitter_type_e emitter_type_t;
enum emitter_type_e { enum emitter_type_e {
emitter_type_bool, emitter_type_bool,
emitter_type_int, emitter_type_int,
emitter_type_int64,
emitter_type_unsigned, emitter_type_unsigned,
emitter_type_uint32, emitter_type_uint32,
emitter_type_uint64, emitter_type_uint64,
...@@ -66,7 +68,7 @@ typedef struct emitter_s emitter_t; ...@@ -66,7 +68,7 @@ typedef struct emitter_s emitter_t;
struct emitter_s { struct emitter_s {
emitter_output_t output; emitter_output_t output;
/* The output information. */ /* The output information. */
void (*write_cb)(void *, const char *); write_cb_t *write_cb;
void *cbopaque; void *cbopaque;
int nesting_depth; int nesting_depth;
/* True if we've already emitted a value at the given depth. */ /* True if we've already emitted a value at the given depth. */
...@@ -75,6 +77,12 @@ struct emitter_s { ...@@ -75,6 +77,12 @@ struct emitter_s {
bool emitted_key; bool emitted_key;
}; };
static inline bool
emitter_outputs_json(emitter_t *emitter) {
return emitter->output == emitter_output_json ||
emitter->output == emitter_output_json_compact;
}
/* Internal convenience function. Write to the emitter the given string. */ /* Internal convenience function. Write to the emitter the given string. */
JEMALLOC_FORMAT_PRINTF(2, 3) JEMALLOC_FORMAT_PRINTF(2, 3)
static inline void static inline void
...@@ -142,6 +150,9 @@ emitter_print_value(emitter_t *emitter, emitter_justify_t justify, int width, ...@@ -142,6 +150,9 @@ emitter_print_value(emitter_t *emitter, emitter_justify_t justify, int width,
case emitter_type_int: case emitter_type_int:
EMIT_SIMPLE(int, "%d") EMIT_SIMPLE(int, "%d")
break; break;
case emitter_type_int64:
EMIT_SIMPLE(int64_t, "%" FMTd64)
break;
case emitter_type_unsigned: case emitter_type_unsigned:
EMIT_SIMPLE(unsigned, "%u") EMIT_SIMPLE(unsigned, "%u")
break; break;
...@@ -196,6 +207,7 @@ static inline void ...@@ -196,6 +207,7 @@ static inline void
emitter_indent(emitter_t *emitter) { emitter_indent(emitter_t *emitter) {
int amount = emitter->nesting_depth; int amount = emitter->nesting_depth;
const char *indent_str; const char *indent_str;
assert(emitter->output != emitter_output_json_compact);
if (emitter->output == emitter_output_json) { if (emitter->output == emitter_output_json) {
indent_str = "\t"; indent_str = "\t";
} else { } else {
...@@ -209,12 +221,18 @@ emitter_indent(emitter_t *emitter) { ...@@ -209,12 +221,18 @@ emitter_indent(emitter_t *emitter) {
static inline void static inline void
emitter_json_key_prefix(emitter_t *emitter) { emitter_json_key_prefix(emitter_t *emitter) {
assert(emitter_outputs_json(emitter));
if (emitter->emitted_key) { if (emitter->emitted_key) {
emitter->emitted_key = false; emitter->emitted_key = false;
return; return;
} }
emitter_printf(emitter, "%s\n", emitter->item_at_depth ? "," : ""); if (emitter->item_at_depth) {
emitter_printf(emitter, ",");
}
if (emitter->output != emitter_output_json_compact) {
emitter_printf(emitter, "\n");
emitter_indent(emitter); emitter_indent(emitter);
}
} }
/******************************************************************************/ /******************************************************************************/
...@@ -222,7 +240,7 @@ emitter_json_key_prefix(emitter_t *emitter) { ...@@ -222,7 +240,7 @@ emitter_json_key_prefix(emitter_t *emitter) {
static inline void static inline void
emitter_init(emitter_t *emitter, emitter_output_t emitter_output, emitter_init(emitter_t *emitter, emitter_output_t emitter_output,
void (*write_cb)(void *, const char *), void *cbopaque) { write_cb_t *write_cb, void *cbopaque) {
emitter->output = emitter_output; emitter->output = emitter_output;
emitter->write_cb = write_cb; emitter->write_cb = write_cb;
emitter->cbopaque = cbopaque; emitter->cbopaque = cbopaque;
...@@ -240,9 +258,10 @@ emitter_init(emitter_t *emitter, emitter_output_t emitter_output, ...@@ -240,9 +258,10 @@ emitter_init(emitter_t *emitter, emitter_output_t emitter_output,
*/ */
static inline void static inline void
emitter_json_key(emitter_t *emitter, const char *json_key) { emitter_json_key(emitter_t *emitter, const char *json_key) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
emitter_json_key_prefix(emitter); emitter_json_key_prefix(emitter);
emitter_printf(emitter, "\"%s\": ", json_key); emitter_printf(emitter, "\"%s\":%s", json_key,
emitter->output == emitter_output_json_compact ? "" : " ");
emitter->emitted_key = true; emitter->emitted_key = true;
} }
} }
...@@ -250,7 +269,7 @@ emitter_json_key(emitter_t *emitter, const char *json_key) { ...@@ -250,7 +269,7 @@ emitter_json_key(emitter_t *emitter, const char *json_key) {
static inline void static inline void
emitter_json_value(emitter_t *emitter, emitter_type_t value_type, emitter_json_value(emitter_t *emitter, emitter_type_t value_type,
const void *value) { const void *value) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
emitter_json_key_prefix(emitter); emitter_json_key_prefix(emitter);
emitter_print_value(emitter, emitter_justify_none, -1, emitter_print_value(emitter, emitter_justify_none, -1,
value_type, value); value_type, value);
...@@ -268,7 +287,7 @@ emitter_json_kv(emitter_t *emitter, const char *json_key, ...@@ -268,7 +287,7 @@ emitter_json_kv(emitter_t *emitter, const char *json_key,
static inline void static inline void
emitter_json_array_begin(emitter_t *emitter) { emitter_json_array_begin(emitter_t *emitter) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
emitter_json_key_prefix(emitter); emitter_json_key_prefix(emitter);
emitter_printf(emitter, "["); emitter_printf(emitter, "[");
emitter_nest_inc(emitter); emitter_nest_inc(emitter);
...@@ -284,18 +303,20 @@ emitter_json_array_kv_begin(emitter_t *emitter, const char *json_key) { ...@@ -284,18 +303,20 @@ emitter_json_array_kv_begin(emitter_t *emitter, const char *json_key) {
static inline void static inline void
emitter_json_array_end(emitter_t *emitter) { emitter_json_array_end(emitter_t *emitter) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
assert(emitter->nesting_depth > 0); assert(emitter->nesting_depth > 0);
emitter_nest_dec(emitter); emitter_nest_dec(emitter);
if (emitter->output != emitter_output_json_compact) {
emitter_printf(emitter, "\n"); emitter_printf(emitter, "\n");
emitter_indent(emitter); emitter_indent(emitter);
}
emitter_printf(emitter, "]"); emitter_printf(emitter, "]");
} }
} }
static inline void static inline void
emitter_json_object_begin(emitter_t *emitter) { emitter_json_object_begin(emitter_t *emitter) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
emitter_json_key_prefix(emitter); emitter_json_key_prefix(emitter);
emitter_printf(emitter, "{"); emitter_printf(emitter, "{");
emitter_nest_inc(emitter); emitter_nest_inc(emitter);
...@@ -311,11 +332,13 @@ emitter_json_object_kv_begin(emitter_t *emitter, const char *json_key) { ...@@ -311,11 +332,13 @@ emitter_json_object_kv_begin(emitter_t *emitter, const char *json_key) {
static inline void static inline void
emitter_json_object_end(emitter_t *emitter) { emitter_json_object_end(emitter_t *emitter) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
assert(emitter->nesting_depth > 0); assert(emitter->nesting_depth > 0);
emitter_nest_dec(emitter); emitter_nest_dec(emitter);
if (emitter->output != emitter_output_json_compact) {
emitter_printf(emitter, "\n"); emitter_printf(emitter, "\n");
emitter_indent(emitter); emitter_indent(emitter);
}
emitter_printf(emitter, "}"); emitter_printf(emitter, "}");
} }
} }
...@@ -420,7 +443,7 @@ emitter_kv_note(emitter_t *emitter, const char *json_key, const char *table_key, ...@@ -420,7 +443,7 @@ emitter_kv_note(emitter_t *emitter, const char *json_key, const char *table_key,
emitter_type_t value_type, const void *value, emitter_type_t value_type, const void *value,
const char *table_note_key, emitter_type_t table_note_value_type, const char *table_note_key, emitter_type_t table_note_value_type,
const void *table_note_value) { const void *table_note_value) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
emitter_json_key(emitter, json_key); emitter_json_key(emitter, json_key);
emitter_json_value(emitter, value_type, value); emitter_json_value(emitter, value_type, value);
} else { } else {
...@@ -440,7 +463,7 @@ emitter_kv(emitter_t *emitter, const char *json_key, const char *table_key, ...@@ -440,7 +463,7 @@ emitter_kv(emitter_t *emitter, const char *json_key, const char *table_key,
static inline void static inline void
emitter_dict_begin(emitter_t *emitter, const char *json_key, emitter_dict_begin(emitter_t *emitter, const char *json_key,
const char *table_header) { const char *table_header) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
emitter_json_key(emitter, json_key); emitter_json_key(emitter, json_key);
emitter_json_object_begin(emitter); emitter_json_object_begin(emitter);
} else { } else {
...@@ -450,7 +473,7 @@ emitter_dict_begin(emitter_t *emitter, const char *json_key, ...@@ -450,7 +473,7 @@ emitter_dict_begin(emitter_t *emitter, const char *json_key,
static inline void static inline void
emitter_dict_end(emitter_t *emitter) { emitter_dict_end(emitter_t *emitter) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
emitter_json_object_end(emitter); emitter_json_object_end(emitter);
} else { } else {
emitter_table_dict_end(emitter); emitter_table_dict_end(emitter);
...@@ -459,7 +482,7 @@ emitter_dict_end(emitter_t *emitter) { ...@@ -459,7 +482,7 @@ emitter_dict_end(emitter_t *emitter) {
static inline void static inline void
emitter_begin(emitter_t *emitter) { emitter_begin(emitter_t *emitter) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
assert(emitter->nesting_depth == 0); assert(emitter->nesting_depth == 0);
emitter_printf(emitter, "{"); emitter_printf(emitter, "{");
emitter_nest_inc(emitter); emitter_nest_inc(emitter);
...@@ -476,10 +499,11 @@ emitter_begin(emitter_t *emitter) { ...@@ -476,10 +499,11 @@ emitter_begin(emitter_t *emitter) {
static inline void static inline void
emitter_end(emitter_t *emitter) { emitter_end(emitter_t *emitter) {
if (emitter->output == emitter_output_json) { if (emitter_outputs_json(emitter)) {
assert(emitter->nesting_depth == 1); assert(emitter->nesting_depth == 1);
emitter_nest_dec(emitter); emitter_nest_dec(emitter);
emitter_printf(emitter, "\n}\n"); emitter_printf(emitter, "%s", emitter->output ==
emitter_output_json_compact ? "}" : "\n}\n");
} }
} }
......
#ifndef JEMALLOC_INTERNAL_ESET_H
#define JEMALLOC_INTERNAL_ESET_H
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/fb.h"
#include "jemalloc/internal/edata.h"
#include "jemalloc/internal/mutex.h"
/*
* An eset ("extent set") is a quantized collection of extents, with built-in
* LRU queue.
*
* This class is not thread-safe; synchronization must be done externally if
* there are mutating operations. One exception is the stats counters, which
* may be read without any locking.
*/
typedef struct eset_bin_s eset_bin_t;
struct eset_bin_s {
edata_heap_t heap;
/*
* We do first-fit across multiple size classes. If we compared against
* the min element in each heap directly, we'd take a cache miss per
* extent we looked at. If we co-locate the edata summaries, we only
* take a miss on the edata we're actually going to return (which is
* inevitable anyways).
*/
edata_cmp_summary_t heap_min;
};
typedef struct eset_bin_stats_s eset_bin_stats_t;
struct eset_bin_stats_s {
atomic_zu_t nextents;
atomic_zu_t nbytes;
};
typedef struct eset_s eset_t;
struct eset_s {
/* Bitmap for which set bits correspond to non-empty heaps. */
fb_group_t bitmap[FB_NGROUPS(SC_NPSIZES + 1)];
/* Quantized per size class heaps of extents. */
eset_bin_t bins[SC_NPSIZES + 1];
eset_bin_stats_t bin_stats[SC_NPSIZES + 1];
/* LRU of all extents in heaps. */
edata_list_inactive_t lru;
/* Page sum for all extents in heaps. */
atomic_zu_t npages;
/*
* A duplication of the data in the containing ecache. We use this only
* for assertions on the states of the passed-in extents.
*/
extent_state_t state;
};
void eset_init(eset_t *eset, extent_state_t state);
size_t eset_npages_get(eset_t *eset);
/* Get the number of extents in the given page size index. */
size_t eset_nextents_get(eset_t *eset, pszind_t ind);
/* Get the sum total bytes of the extents in the given page size index. */
size_t eset_nbytes_get(eset_t *eset, pszind_t ind);
void eset_insert(eset_t *eset, edata_t *edata);
void eset_remove(eset_t *eset, edata_t *edata);
/*
* Select an extent from this eset of the given size and alignment. Returns
* null if no such item could be found.
*/
edata_t *eset_fit(eset_t *eset, size_t esize, size_t alignment, bool exact_only,
unsigned lg_max_fit);
#endif /* JEMALLOC_INTERNAL_ESET_H */
#ifndef JEMALLOC_INTERNAL_EXP_GROW_H
#define JEMALLOC_INTERNAL_EXP_GROW_H
typedef struct exp_grow_s exp_grow_t;
struct exp_grow_s {
/*
* Next extent size class in a growing series to use when satisfying a
* request via the extent hooks (only if opt_retain). This limits the
* number of disjoint virtual memory ranges so that extent merging can
* be effective even if multiple arenas' extent allocation requests are
* highly interleaved.
*
* retain_grow_limit is the max allowed size ind to expand (unless the
* required size is greater). Default is no limit, and controlled
* through mallctl only.
*/
pszind_t next;
pszind_t limit;
};
static inline bool
exp_grow_size_prepare(exp_grow_t *exp_grow, size_t alloc_size_min,
size_t *r_alloc_size, pszind_t *r_skip) {
*r_skip = 0;
*r_alloc_size = sz_pind2sz(exp_grow->next + *r_skip);
while (*r_alloc_size < alloc_size_min) {
(*r_skip)++;
if (exp_grow->next + *r_skip >=
sz_psz2ind(SC_LARGE_MAXCLASS)) {
/* Outside legal range. */
return true;
}
*r_alloc_size = sz_pind2sz(exp_grow->next + *r_skip);
}
return false;
}
static inline void
exp_grow_size_commit(exp_grow_t *exp_grow, pszind_t skip) {
if (exp_grow->next + skip + 1 <= exp_grow->limit) {
exp_grow->next += skip + 1;
} else {
exp_grow->next = exp_grow->limit;
}
}
void exp_grow_init(exp_grow_t *exp_grow);
#endif /* JEMALLOC_INTERNAL_EXP_GROW_H */
#ifndef JEMALLOC_INTERNAL_EXTENT_H
#define JEMALLOC_INTERNAL_EXTENT_H
#include "jemalloc/internal/ecache.h"
#include "jemalloc/internal/ehooks.h"
#include "jemalloc/internal/ph.h"
#include "jemalloc/internal/rtree.h"
/*
* This module contains the page-level allocator. It chooses the addresses that
* allocations requested by other modules will inhabit, and updates the global
* metadata to reflect allocation/deallocation/purging decisions.
*/
/*
* When reuse (and split) an active extent, (1U << opt_lg_extent_max_active_fit)
* is the max ratio between the size of the active extent and the new extent.
*/
#define LG_EXTENT_MAX_ACTIVE_FIT_DEFAULT 6
extern size_t opt_lg_extent_max_active_fit;
edata_t *ecache_alloc(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, edata_t *expand_edata, size_t size, size_t alignment,
bool zero, bool guarded);
edata_t *ecache_alloc_grow(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, edata_t *expand_edata, size_t size, size_t alignment,
bool zero, bool guarded);
void ecache_dalloc(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, edata_t *edata);
edata_t *ecache_evict(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, size_t npages_min);
void extent_gdump_add(tsdn_t *tsdn, const edata_t *edata);
void extent_record(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, ecache_t *ecache,
edata_t *edata);
void extent_dalloc_gap(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *edata);
edata_t *extent_alloc_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
void *new_addr, size_t size, size_t alignment, bool zero, bool *commit,
bool growing_retained);
void extent_dalloc_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *edata);
void extent_destroy_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *edata);
bool extent_commit_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
bool extent_decommit_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
bool extent_purge_lazy_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
bool extent_purge_forced_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
edata_t *extent_split_wrapper(tsdn_t *tsdn, pac_t *pac,
ehooks_t *ehooks, edata_t *edata, size_t size_a, size_t size_b,
bool holding_core_locks);
bool extent_merge_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *a, edata_t *b);
bool extent_commit_zero(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
bool commit, bool zero, bool growing_retained);
size_t extent_sn_next(pac_t *pac);
bool extent_boot(void);
JEMALLOC_ALWAYS_INLINE bool
extent_neighbor_head_state_mergeable(bool edata_is_head,
bool neighbor_is_head, bool forward) {
/*
* Head states checking: disallow merging if the higher addr extent is a
* head extent. This helps preserve first-fit, and more importantly
* makes sure no merge across arenas.
*/
if (forward) {
if (neighbor_is_head) {
return false;
}
} else {
if (edata_is_head) {
return false;
}
}
return true;
}
JEMALLOC_ALWAYS_INLINE bool
extent_can_acquire_neighbor(edata_t *edata, rtree_contents_t contents,
extent_pai_t pai, extent_state_t expected_state, bool forward,
bool expanding) {
edata_t *neighbor = contents.edata;
if (neighbor == NULL) {
return false;
}
/* It's not safe to access *neighbor yet; must verify states first. */
bool neighbor_is_head = contents.metadata.is_head;
if (!extent_neighbor_head_state_mergeable(edata_is_head_get(edata),
neighbor_is_head, forward)) {
return false;
}
extent_state_t neighbor_state = contents.metadata.state;
if (pai == EXTENT_PAI_PAC) {
if (neighbor_state != expected_state) {
return false;
}
/* From this point, it's safe to access *neighbor. */
if (!expanding && (edata_committed_get(edata) !=
edata_committed_get(neighbor))) {
/*
* Some platforms (e.g. Windows) require an explicit
* commit step (and writing to uncommitted memory is not
* allowed).
*/
return false;
}
} else {
if (neighbor_state == extent_state_active) {
return false;
}
/* From this point, it's safe to access *neighbor. */
}
assert(edata_pai_get(edata) == pai);
if (edata_pai_get(neighbor) != pai) {
return false;
}
if (opt_retain) {
assert(edata_arena_ind_get(edata) ==
edata_arena_ind_get(neighbor));
} else {
if (edata_arena_ind_get(edata) !=
edata_arena_ind_get(neighbor)) {
return false;
}
}
assert(!edata_guarded_get(edata) && !edata_guarded_get(neighbor));
return true;
}
#endif /* JEMALLOC_INTERNAL_EXTENT_H */
#ifndef JEMALLOC_INTERNAL_EXTENT_EXTERNS_H
#define JEMALLOC_INTERNAL_EXTENT_EXTERNS_H
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/mutex_pool.h"
#include "jemalloc/internal/ph.h"
#include "jemalloc/internal/rtree.h"
extern size_t opt_lg_extent_max_active_fit;
extern rtree_t extents_rtree;
extern const extent_hooks_t extent_hooks_default;
extern mutex_pool_t extent_mutex_pool;
extent_t *extent_alloc(tsdn_t *tsdn, arena_t *arena);
void extent_dalloc(tsdn_t *tsdn, arena_t *arena, extent_t *extent);
extent_hooks_t *extent_hooks_get(arena_t *arena);
extent_hooks_t *extent_hooks_set(tsd_t *tsd, arena_t *arena,
extent_hooks_t *extent_hooks);
#ifdef JEMALLOC_JET
size_t extent_size_quantize_floor(size_t size);
size_t extent_size_quantize_ceil(size_t size);
#endif
ph_proto(, extent_avail_, extent_tree_t, extent_t)
ph_proto(, extent_heap_, extent_heap_t, extent_t)
bool extents_init(tsdn_t *tsdn, extents_t *extents, extent_state_t state,
bool delay_coalesce);
extent_state_t extents_state_get(const extents_t *extents);
size_t extents_npages_get(extents_t *extents);
/* Get the number of extents in the given page size index. */
size_t extents_nextents_get(extents_t *extents, pszind_t ind);
/* Get the sum total bytes of the extents in the given page size index. */
size_t extents_nbytes_get(extents_t *extents, pszind_t ind);
extent_t *extents_alloc(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extents_t *extents, void *new_addr,
size_t size, size_t pad, size_t alignment, bool slab, szind_t szind,
bool *zero, bool *commit);
void extents_dalloc(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extents_t *extents, extent_t *extent);
extent_t *extents_evict(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extents_t *extents, size_t npages_min);
void extents_prefork(tsdn_t *tsdn, extents_t *extents);
void extents_postfork_parent(tsdn_t *tsdn, extents_t *extents);
void extents_postfork_child(tsdn_t *tsdn, extents_t *extents);
extent_t *extent_alloc_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, void *new_addr, size_t size, size_t pad,
size_t alignment, bool slab, szind_t szind, bool *zero, bool *commit);
void extent_dalloc_gap(tsdn_t *tsdn, arena_t *arena, extent_t *extent);
void extent_dalloc_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *extent);
void extent_destroy_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *extent);
bool extent_commit_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
size_t length);
bool extent_decommit_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
size_t length);
bool extent_purge_lazy_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
size_t length);
bool extent_purge_forced_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *extent, size_t offset,
size_t length);
extent_t *extent_split_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *extent, size_t size_a,
szind_t szind_a, bool slab_a, size_t size_b, szind_t szind_b, bool slab_b);
bool extent_merge_wrapper(tsdn_t *tsdn, arena_t *arena,
extent_hooks_t **r_extent_hooks, extent_t *a, extent_t *b);
bool extent_boot(void);
void extent_util_stats_get(tsdn_t *tsdn, const void *ptr,
size_t *nfree, size_t *nregs, size_t *size);
void extent_util_stats_verbose_get(tsdn_t *tsdn, const void *ptr,
size_t *nfree, size_t *nregs, size_t *size,
size_t *bin_nfree, size_t *bin_nregs, void **slabcur_addr);
#endif /* JEMALLOC_INTERNAL_EXTENT_EXTERNS_H */
#ifndef JEMALLOC_INTERNAL_EXTENT_INLINES_H
#define JEMALLOC_INTERNAL_EXTENT_INLINES_H
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/mutex_pool.h"
#include "jemalloc/internal/pages.h"
#include "jemalloc/internal/prng.h"
#include "jemalloc/internal/ql.h"
#include "jemalloc/internal/sc.h"
#include "jemalloc/internal/sz.h"
static inline void
extent_lock(tsdn_t *tsdn, extent_t *extent) {
assert(extent != NULL);
mutex_pool_lock(tsdn, &extent_mutex_pool, (uintptr_t)extent);
}
static inline void
extent_unlock(tsdn_t *tsdn, extent_t *extent) {
assert(extent != NULL);
mutex_pool_unlock(tsdn, &extent_mutex_pool, (uintptr_t)extent);
}
static inline void
extent_lock2(tsdn_t *tsdn, extent_t *extent1, extent_t *extent2) {
assert(extent1 != NULL && extent2 != NULL);
mutex_pool_lock2(tsdn, &extent_mutex_pool, (uintptr_t)extent1,
(uintptr_t)extent2);
}
static inline void
extent_unlock2(tsdn_t *tsdn, extent_t *extent1, extent_t *extent2) {
assert(extent1 != NULL && extent2 != NULL);
mutex_pool_unlock2(tsdn, &extent_mutex_pool, (uintptr_t)extent1,
(uintptr_t)extent2);
}
static inline unsigned
extent_arena_ind_get(const extent_t *extent) {
unsigned arena_ind = (unsigned)((extent->e_bits &
EXTENT_BITS_ARENA_MASK) >> EXTENT_BITS_ARENA_SHIFT);
assert(arena_ind < MALLOCX_ARENA_LIMIT);
return arena_ind;
}
static inline arena_t *
extent_arena_get(const extent_t *extent) {
unsigned arena_ind = extent_arena_ind_get(extent);
return (arena_t *)atomic_load_p(&arenas[arena_ind], ATOMIC_ACQUIRE);
}
static inline szind_t
extent_szind_get_maybe_invalid(const extent_t *extent) {
szind_t szind = (szind_t)((extent->e_bits & EXTENT_BITS_SZIND_MASK) >>
EXTENT_BITS_SZIND_SHIFT);
assert(szind <= SC_NSIZES);
return szind;
}
static inline szind_t
extent_szind_get(const extent_t *extent) {
szind_t szind = extent_szind_get_maybe_invalid(extent);
assert(szind < SC_NSIZES); /* Never call when "invalid". */
return szind;
}
static inline size_t
extent_usize_get(const extent_t *extent) {
return sz_index2size(extent_szind_get(extent));
}
static inline unsigned
extent_binshard_get(const extent_t *extent) {
unsigned binshard = (unsigned)((extent->e_bits &
EXTENT_BITS_BINSHARD_MASK) >> EXTENT_BITS_BINSHARD_SHIFT);
assert(binshard < bin_infos[extent_szind_get(extent)].n_shards);
return binshard;
}
static inline size_t
extent_sn_get(const extent_t *extent) {
return (size_t)((extent->e_bits & EXTENT_BITS_SN_MASK) >>
EXTENT_BITS_SN_SHIFT);
}
static inline extent_state_t
extent_state_get(const extent_t *extent) {
return (extent_state_t)((extent->e_bits & EXTENT_BITS_STATE_MASK) >>
EXTENT_BITS_STATE_SHIFT);
}
static inline bool
extent_zeroed_get(const extent_t *extent) {
return (bool)((extent->e_bits & EXTENT_BITS_ZEROED_MASK) >>
EXTENT_BITS_ZEROED_SHIFT);
}
static inline bool
extent_committed_get(const extent_t *extent) {
return (bool)((extent->e_bits & EXTENT_BITS_COMMITTED_MASK) >>
EXTENT_BITS_COMMITTED_SHIFT);
}
static inline bool
extent_dumpable_get(const extent_t *extent) {
return (bool)((extent->e_bits & EXTENT_BITS_DUMPABLE_MASK) >>
EXTENT_BITS_DUMPABLE_SHIFT);
}
static inline bool
extent_slab_get(const extent_t *extent) {
return (bool)((extent->e_bits & EXTENT_BITS_SLAB_MASK) >>
EXTENT_BITS_SLAB_SHIFT);
}
static inline unsigned
extent_nfree_get(const extent_t *extent) {
assert(extent_slab_get(extent));
return (unsigned)((extent->e_bits & EXTENT_BITS_NFREE_MASK) >>
EXTENT_BITS_NFREE_SHIFT);
}
static inline void *
extent_base_get(const extent_t *extent) {
assert(extent->e_addr == PAGE_ADDR2BASE(extent->e_addr) ||
!extent_slab_get(extent));
return PAGE_ADDR2BASE(extent->e_addr);
}
static inline void *
extent_addr_get(const extent_t *extent) {
assert(extent->e_addr == PAGE_ADDR2BASE(extent->e_addr) ||
!extent_slab_get(extent));
return extent->e_addr;
}
static inline size_t
extent_size_get(const extent_t *extent) {
return (extent->e_size_esn & EXTENT_SIZE_MASK);
}
static inline size_t
extent_esn_get(const extent_t *extent) {
return (extent->e_size_esn & EXTENT_ESN_MASK);
}
static inline size_t
extent_bsize_get(const extent_t *extent) {
return extent->e_bsize;
}
static inline void *
extent_before_get(const extent_t *extent) {
return (void *)((uintptr_t)extent_base_get(extent) - PAGE);
}
static inline void *
extent_last_get(const extent_t *extent) {
return (void *)((uintptr_t)extent_base_get(extent) +
extent_size_get(extent) - PAGE);
}
static inline void *
extent_past_get(const extent_t *extent) {
return (void *)((uintptr_t)extent_base_get(extent) +
extent_size_get(extent));
}
static inline arena_slab_data_t *
extent_slab_data_get(extent_t *extent) {
assert(extent_slab_get(extent));
return &extent->e_slab_data;
}
static inline const arena_slab_data_t *
extent_slab_data_get_const(const extent_t *extent) {
assert(extent_slab_get(extent));
return &extent->e_slab_data;
}
static inline prof_tctx_t *
extent_prof_tctx_get(const extent_t *extent) {
return (prof_tctx_t *)atomic_load_p(&extent->e_prof_tctx,
ATOMIC_ACQUIRE);
}
static inline nstime_t
extent_prof_alloc_time_get(const extent_t *extent) {
return extent->e_alloc_time;
}
static inline void
extent_arena_set(extent_t *extent, arena_t *arena) {
unsigned arena_ind = (arena != NULL) ? arena_ind_get(arena) : ((1U <<
MALLOCX_ARENA_BITS) - 1);
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_ARENA_MASK) |
((uint64_t)arena_ind << EXTENT_BITS_ARENA_SHIFT);
}
static inline void
extent_binshard_set(extent_t *extent, unsigned binshard) {
/* The assertion assumes szind is set already. */
assert(binshard < bin_infos[extent_szind_get(extent)].n_shards);
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_BINSHARD_MASK) |
((uint64_t)binshard << EXTENT_BITS_BINSHARD_SHIFT);
}
static inline void
extent_addr_set(extent_t *extent, void *addr) {
extent->e_addr = addr;
}
static inline void
extent_addr_randomize(tsdn_t *tsdn, extent_t *extent, size_t alignment) {
assert(extent_base_get(extent) == extent_addr_get(extent));
if (alignment < PAGE) {
unsigned lg_range = LG_PAGE -
lg_floor(CACHELINE_CEILING(alignment));
size_t r;
if (!tsdn_null(tsdn)) {
tsd_t *tsd = tsdn_tsd(tsdn);
r = (size_t)prng_lg_range_u64(
tsd_offset_statep_get(tsd), lg_range);
} else {
r = prng_lg_range_zu(
&extent_arena_get(extent)->offset_state,
lg_range, true);
}
uintptr_t random_offset = ((uintptr_t)r) << (LG_PAGE -
lg_range);
extent->e_addr = (void *)((uintptr_t)extent->e_addr +
random_offset);
assert(ALIGNMENT_ADDR2BASE(extent->e_addr, alignment) ==
extent->e_addr);
}
}
static inline void
extent_size_set(extent_t *extent, size_t size) {
assert((size & ~EXTENT_SIZE_MASK) == 0);
extent->e_size_esn = size | (extent->e_size_esn & ~EXTENT_SIZE_MASK);
}
static inline void
extent_esn_set(extent_t *extent, size_t esn) {
extent->e_size_esn = (extent->e_size_esn & ~EXTENT_ESN_MASK) | (esn &
EXTENT_ESN_MASK);
}
static inline void
extent_bsize_set(extent_t *extent, size_t bsize) {
extent->e_bsize = bsize;
}
static inline void
extent_szind_set(extent_t *extent, szind_t szind) {
assert(szind <= SC_NSIZES); /* SC_NSIZES means "invalid". */
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_SZIND_MASK) |
((uint64_t)szind << EXTENT_BITS_SZIND_SHIFT);
}
static inline void
extent_nfree_set(extent_t *extent, unsigned nfree) {
assert(extent_slab_get(extent));
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_NFREE_MASK) |
((uint64_t)nfree << EXTENT_BITS_NFREE_SHIFT);
}
static inline void
extent_nfree_binshard_set(extent_t *extent, unsigned nfree, unsigned binshard) {
/* The assertion assumes szind is set already. */
assert(binshard < bin_infos[extent_szind_get(extent)].n_shards);
extent->e_bits = (extent->e_bits &
(~EXTENT_BITS_NFREE_MASK & ~EXTENT_BITS_BINSHARD_MASK)) |
((uint64_t)binshard << EXTENT_BITS_BINSHARD_SHIFT) |
((uint64_t)nfree << EXTENT_BITS_NFREE_SHIFT);
}
static inline void
extent_nfree_inc(extent_t *extent) {
assert(extent_slab_get(extent));
extent->e_bits += ((uint64_t)1U << EXTENT_BITS_NFREE_SHIFT);
}
static inline void
extent_nfree_dec(extent_t *extent) {
assert(extent_slab_get(extent));
extent->e_bits -= ((uint64_t)1U << EXTENT_BITS_NFREE_SHIFT);
}
static inline void
extent_nfree_sub(extent_t *extent, uint64_t n) {
assert(extent_slab_get(extent));
extent->e_bits -= (n << EXTENT_BITS_NFREE_SHIFT);
}
static inline void
extent_sn_set(extent_t *extent, size_t sn) {
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_SN_MASK) |
((uint64_t)sn << EXTENT_BITS_SN_SHIFT);
}
static inline void
extent_state_set(extent_t *extent, extent_state_t state) {
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_STATE_MASK) |
((uint64_t)state << EXTENT_BITS_STATE_SHIFT);
}
static inline void
extent_zeroed_set(extent_t *extent, bool zeroed) {
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_ZEROED_MASK) |
((uint64_t)zeroed << EXTENT_BITS_ZEROED_SHIFT);
}
static inline void
extent_committed_set(extent_t *extent, bool committed) {
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_COMMITTED_MASK) |
((uint64_t)committed << EXTENT_BITS_COMMITTED_SHIFT);
}
static inline void
extent_dumpable_set(extent_t *extent, bool dumpable) {
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_DUMPABLE_MASK) |
((uint64_t)dumpable << EXTENT_BITS_DUMPABLE_SHIFT);
}
static inline void
extent_slab_set(extent_t *extent, bool slab) {
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_SLAB_MASK) |
((uint64_t)slab << EXTENT_BITS_SLAB_SHIFT);
}
static inline void
extent_prof_tctx_set(extent_t *extent, prof_tctx_t *tctx) {
atomic_store_p(&extent->e_prof_tctx, tctx, ATOMIC_RELEASE);
}
static inline void
extent_prof_alloc_time_set(extent_t *extent, nstime_t t) {
nstime_copy(&extent->e_alloc_time, &t);
}
static inline bool
extent_is_head_get(extent_t *extent) {
if (maps_coalesce) {
not_reached();
}
return (bool)((extent->e_bits & EXTENT_BITS_IS_HEAD_MASK) >>
EXTENT_BITS_IS_HEAD_SHIFT);
}
static inline void
extent_is_head_set(extent_t *extent, bool is_head) {
if (maps_coalesce) {
not_reached();
}
extent->e_bits = (extent->e_bits & ~EXTENT_BITS_IS_HEAD_MASK) |
((uint64_t)is_head << EXTENT_BITS_IS_HEAD_SHIFT);
}
static inline void
extent_init(extent_t *extent, arena_t *arena, void *addr, size_t size,
bool slab, szind_t szind, size_t sn, extent_state_t state, bool zeroed,
bool committed, bool dumpable, extent_head_state_t is_head) {
assert(addr == PAGE_ADDR2BASE(addr) || !slab);
extent_arena_set(extent, arena);
extent_addr_set(extent, addr);
extent_size_set(extent, size);
extent_slab_set(extent, slab);
extent_szind_set(extent, szind);
extent_sn_set(extent, sn);
extent_state_set(extent, state);
extent_zeroed_set(extent, zeroed);
extent_committed_set(extent, committed);
extent_dumpable_set(extent, dumpable);
ql_elm_new(extent, ql_link);
if (!maps_coalesce) {
extent_is_head_set(extent, (is_head == EXTENT_IS_HEAD) ? true :
false);
}
if (config_prof) {
extent_prof_tctx_set(extent, NULL);
}
}
static inline void
extent_binit(extent_t *extent, void *addr, size_t bsize, size_t sn) {
extent_arena_set(extent, NULL);
extent_addr_set(extent, addr);
extent_bsize_set(extent, bsize);
extent_slab_set(extent, false);
extent_szind_set(extent, SC_NSIZES);
extent_sn_set(extent, sn);
extent_state_set(extent, extent_state_active);
extent_zeroed_set(extent, true);
extent_committed_set(extent, true);
extent_dumpable_set(extent, true);
}
static inline void
extent_list_init(extent_list_t *list) {
ql_new(list);
}
static inline extent_t *
extent_list_first(const extent_list_t *list) {
return ql_first(list);
}
static inline extent_t *
extent_list_last(const extent_list_t *list) {
return ql_last(list, ql_link);
}
static inline void
extent_list_append(extent_list_t *list, extent_t *extent) {
ql_tail_insert(list, extent, ql_link);
}
static inline void
extent_list_prepend(extent_list_t *list, extent_t *extent) {
ql_head_insert(list, extent, ql_link);
}
static inline void
extent_list_replace(extent_list_t *list, extent_t *to_remove,
extent_t *to_insert) {
ql_after_insert(to_remove, to_insert, ql_link);
ql_remove(list, to_remove, ql_link);
}
static inline void
extent_list_remove(extent_list_t *list, extent_t *extent) {
ql_remove(list, extent, ql_link);
}
static inline int
extent_sn_comp(const extent_t *a, const extent_t *b) {
size_t a_sn = extent_sn_get(a);
size_t b_sn = extent_sn_get(b);
return (a_sn > b_sn) - (a_sn < b_sn);
}
static inline int
extent_esn_comp(const extent_t *a, const extent_t *b) {
size_t a_esn = extent_esn_get(a);
size_t b_esn = extent_esn_get(b);
return (a_esn > b_esn) - (a_esn < b_esn);
}
static inline int
extent_ad_comp(const extent_t *a, const extent_t *b) {
uintptr_t a_addr = (uintptr_t)extent_addr_get(a);
uintptr_t b_addr = (uintptr_t)extent_addr_get(b);
return (a_addr > b_addr) - (a_addr < b_addr);
}
static inline int
extent_ead_comp(const extent_t *a, const extent_t *b) {
uintptr_t a_eaddr = (uintptr_t)a;
uintptr_t b_eaddr = (uintptr_t)b;
return (a_eaddr > b_eaddr) - (a_eaddr < b_eaddr);
}
static inline int
extent_snad_comp(const extent_t *a, const extent_t *b) {
int ret;
ret = extent_sn_comp(a, b);
if (ret != 0) {
return ret;
}
ret = extent_ad_comp(a, b);
return ret;
}
static inline int
extent_esnead_comp(const extent_t *a, const extent_t *b) {
int ret;
ret = extent_esn_comp(a, b);
if (ret != 0) {
return ret;
}
ret = extent_ead_comp(a, b);
return ret;
}
#endif /* JEMALLOC_INTERNAL_EXTENT_INLINES_H */
#ifndef JEMALLOC_INTERNAL_EXTENT_STRUCTS_H
#define JEMALLOC_INTERNAL_EXTENT_STRUCTS_H
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/bit_util.h"
#include "jemalloc/internal/bitmap.h"
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/ql.h"
#include "jemalloc/internal/ph.h"
#include "jemalloc/internal/sc.h"
typedef enum {
extent_state_active = 0,
extent_state_dirty = 1,
extent_state_muzzy = 2,
extent_state_retained = 3
} extent_state_t;
/* Extent (span of pages). Use accessor functions for e_* fields. */
struct extent_s {
/*
* Bitfield containing several fields:
*
* a: arena_ind
* b: slab
* c: committed
* d: dumpable
* z: zeroed
* t: state
* i: szind
* f: nfree
* s: bin_shard
* n: sn
*
* nnnnnnnn ... nnnnnnss ssssffff ffffffii iiiiiitt zdcbaaaa aaaaaaaa
*
* arena_ind: Arena from which this extent came, or all 1 bits if
* unassociated.
*
* slab: The slab flag indicates whether the extent is used for a slab
* of small regions. This helps differentiate small size classes,
* and it indicates whether interior pointers can be looked up via
* iealloc().
*
* committed: The committed flag indicates whether physical memory is
* committed to the extent, whether explicitly or implicitly
* as on a system that overcommits and satisfies physical
* memory needs on demand via soft page faults.
*
* dumpable: The dumpable flag indicates whether or not we've set the
* memory in question to be dumpable. Note that this
* interacts somewhat subtly with user-specified extent hooks,
* since we don't know if *they* are fiddling with
* dumpability (in which case, we don't want to undo whatever
* they're doing). To deal with this scenario, we:
* - Make dumpable false only for memory allocated with the
* default hooks.
* - Only allow memory to go from non-dumpable to dumpable,
* and only once.
* - Never make the OS call to allow dumping when the
* dumpable bit is already set.
* These three constraints mean that we will never
* accidentally dump user memory that the user meant to set
* nondumpable with their extent hooks.
*
*
* zeroed: The zeroed flag is used by extent recycling code to track
* whether memory is zero-filled.
*
* state: The state flag is an extent_state_t.
*
* szind: The szind flag indicates usable size class index for
* allocations residing in this extent, regardless of whether the
* extent is a slab. Extent size and usable size often differ
* even for non-slabs, either due to sz_large_pad or promotion of
* sampled small regions.
*
* nfree: Number of free regions in slab.
*
* bin_shard: the shard of the bin from which this extent came.
*
* sn: Serial number (potentially non-unique).
*
* Serial numbers may wrap around if !opt_retain, but as long as
* comparison functions fall back on address comparison for equal
* serial numbers, stable (if imperfect) ordering is maintained.
*
* Serial numbers may not be unique even in the absence of
* wrap-around, e.g. when splitting an extent and assigning the same
* serial number to both resulting adjacent extents.
*/
uint64_t e_bits;
#define MASK(CURRENT_FIELD_WIDTH, CURRENT_FIELD_SHIFT) ((((((uint64_t)0x1U) << (CURRENT_FIELD_WIDTH)) - 1)) << (CURRENT_FIELD_SHIFT))
#define EXTENT_BITS_ARENA_WIDTH MALLOCX_ARENA_BITS
#define EXTENT_BITS_ARENA_SHIFT 0
#define EXTENT_BITS_ARENA_MASK MASK(EXTENT_BITS_ARENA_WIDTH, EXTENT_BITS_ARENA_SHIFT)
#define EXTENT_BITS_SLAB_WIDTH 1
#define EXTENT_BITS_SLAB_SHIFT (EXTENT_BITS_ARENA_WIDTH + EXTENT_BITS_ARENA_SHIFT)
#define EXTENT_BITS_SLAB_MASK MASK(EXTENT_BITS_SLAB_WIDTH, EXTENT_BITS_SLAB_SHIFT)
#define EXTENT_BITS_COMMITTED_WIDTH 1
#define EXTENT_BITS_COMMITTED_SHIFT (EXTENT_BITS_SLAB_WIDTH + EXTENT_BITS_SLAB_SHIFT)
#define EXTENT_BITS_COMMITTED_MASK MASK(EXTENT_BITS_COMMITTED_WIDTH, EXTENT_BITS_COMMITTED_SHIFT)
#define EXTENT_BITS_DUMPABLE_WIDTH 1
#define EXTENT_BITS_DUMPABLE_SHIFT (EXTENT_BITS_COMMITTED_WIDTH + EXTENT_BITS_COMMITTED_SHIFT)
#define EXTENT_BITS_DUMPABLE_MASK MASK(EXTENT_BITS_DUMPABLE_WIDTH, EXTENT_BITS_DUMPABLE_SHIFT)
#define EXTENT_BITS_ZEROED_WIDTH 1
#define EXTENT_BITS_ZEROED_SHIFT (EXTENT_BITS_DUMPABLE_WIDTH + EXTENT_BITS_DUMPABLE_SHIFT)
#define EXTENT_BITS_ZEROED_MASK MASK(EXTENT_BITS_ZEROED_WIDTH, EXTENT_BITS_ZEROED_SHIFT)
#define EXTENT_BITS_STATE_WIDTH 2
#define EXTENT_BITS_STATE_SHIFT (EXTENT_BITS_ZEROED_WIDTH + EXTENT_BITS_ZEROED_SHIFT)
#define EXTENT_BITS_STATE_MASK MASK(EXTENT_BITS_STATE_WIDTH, EXTENT_BITS_STATE_SHIFT)
#define EXTENT_BITS_SZIND_WIDTH LG_CEIL(SC_NSIZES)
#define EXTENT_BITS_SZIND_SHIFT (EXTENT_BITS_STATE_WIDTH + EXTENT_BITS_STATE_SHIFT)
#define EXTENT_BITS_SZIND_MASK MASK(EXTENT_BITS_SZIND_WIDTH, EXTENT_BITS_SZIND_SHIFT)
#define EXTENT_BITS_NFREE_WIDTH (LG_SLAB_MAXREGS + 1)
#define EXTENT_BITS_NFREE_SHIFT (EXTENT_BITS_SZIND_WIDTH + EXTENT_BITS_SZIND_SHIFT)
#define EXTENT_BITS_NFREE_MASK MASK(EXTENT_BITS_NFREE_WIDTH, EXTENT_BITS_NFREE_SHIFT)
#define EXTENT_BITS_BINSHARD_WIDTH 6
#define EXTENT_BITS_BINSHARD_SHIFT (EXTENT_BITS_NFREE_WIDTH + EXTENT_BITS_NFREE_SHIFT)
#define EXTENT_BITS_BINSHARD_MASK MASK(EXTENT_BITS_BINSHARD_WIDTH, EXTENT_BITS_BINSHARD_SHIFT)
#define EXTENT_BITS_IS_HEAD_WIDTH 1
#define EXTENT_BITS_IS_HEAD_SHIFT (EXTENT_BITS_BINSHARD_WIDTH + EXTENT_BITS_BINSHARD_SHIFT)
#define EXTENT_BITS_IS_HEAD_MASK MASK(EXTENT_BITS_IS_HEAD_WIDTH, EXTENT_BITS_IS_HEAD_SHIFT)
#define EXTENT_BITS_SN_SHIFT (EXTENT_BITS_IS_HEAD_WIDTH + EXTENT_BITS_IS_HEAD_SHIFT)
#define EXTENT_BITS_SN_MASK (UINT64_MAX << EXTENT_BITS_SN_SHIFT)
/* Pointer to the extent that this structure is responsible for. */
void *e_addr;
union {
/*
* Extent size and serial number associated with the extent
* structure (different than the serial number for the extent at
* e_addr).
*
* ssssssss [...] ssssssss ssssnnnn nnnnnnnn
*/
size_t e_size_esn;
#define EXTENT_SIZE_MASK ((size_t)~(PAGE-1))
#define EXTENT_ESN_MASK ((size_t)PAGE-1)
/* Base extent size, which may not be a multiple of PAGE. */
size_t e_bsize;
};
/*
* List linkage, used by a variety of lists:
* - bin_t's slabs_full
* - extents_t's LRU
* - stashed dirty extents
* - arena's large allocations
*/
ql_elm(extent_t) ql_link;
/*
* Linkage for per size class sn/address-ordered heaps, and
* for extent_avail
*/
phn(extent_t) ph_link;
union {
/* Small region slab metadata. */
arena_slab_data_t e_slab_data;
/* Profiling data, used for large objects. */
struct {
/* Time when this was allocated. */
nstime_t e_alloc_time;
/* Points to a prof_tctx_t. */
atomic_p_t e_prof_tctx;
};
};
};
typedef ql_head(extent_t) extent_list_t;
typedef ph(extent_t) extent_tree_t;
typedef ph(extent_t) extent_heap_t;
/* Quantized collection of extents, with built-in LRU queue. */
struct extents_s {
malloc_mutex_t mtx;
/*
* Quantized per size class heaps of extents.
*
* Synchronization: mtx.
*/
extent_heap_t heaps[SC_NPSIZES + 1];
atomic_zu_t nextents[SC_NPSIZES + 1];
atomic_zu_t nbytes[SC_NPSIZES + 1];
/*
* Bitmap for which set bits correspond to non-empty heaps.
*
* Synchronization: mtx.
*/
bitmap_t bitmap[BITMAP_GROUPS(SC_NPSIZES + 1)];
/*
* LRU of all extents in heaps.
*
* Synchronization: mtx.
*/
extent_list_t lru;
/*
* Page sum for all extents in heaps.
*
* The synchronization here is a little tricky. Modifications to npages
* must hold mtx, but reads need not (though, a reader who sees npages
* without holding the mutex can't assume anything about the rest of the
* state of the extents_t).
*/
atomic_zu_t npages;
/* All stored extents must be in the same state. */
extent_state_t state;
/*
* If true, delay coalescing until eviction; otherwise coalesce during
* deallocation.
*/
bool delay_coalesce;
};
/*
* The following two structs are for experimental purposes. See
* experimental_utilization_query_ctl and
* experimental_utilization_batch_query_ctl in src/ctl.c.
*/
struct extent_util_stats_s {
size_t nfree;
size_t nregs;
size_t size;
};
struct extent_util_stats_verbose_s {
void *slabcur_addr;
size_t nfree;
size_t nregs;
size_t size;
size_t bin_nfree;
size_t bin_nregs;
};
#endif /* JEMALLOC_INTERNAL_EXTENT_STRUCTS_H */
#ifndef JEMALLOC_INTERNAL_EXTENT_TYPES_H
#define JEMALLOC_INTERNAL_EXTENT_TYPES_H
typedef struct extent_s extent_t;
typedef struct extents_s extents_t;
typedef struct extent_util_stats_s extent_util_stats_t;
typedef struct extent_util_stats_verbose_s extent_util_stats_verbose_t;
#define EXTENT_HOOKS_INITIALIZER NULL
/*
* When reuse (and split) an active extent, (1U << opt_lg_extent_max_active_fit)
* is the max ratio between the size of the active extent and the new extent.
*/
#define LG_EXTENT_MAX_ACTIVE_FIT_DEFAULT 6
typedef enum {
EXTENT_NOT_HEAD,
EXTENT_IS_HEAD /* Only relevant for Windows && opt.retain. */
} extent_head_state_t;
#endif /* JEMALLOC_INTERNAL_EXTENT_TYPES_H */
#ifndef JEMALLOC_INTERNAL_FB_H
#define JEMALLOC_INTERNAL_FB_H
/*
* The flat bitmap module. This has a larger API relative to the bitmap module
* (supporting things like backwards searches, and searching for both set and
* unset bits), at the cost of slower operations for very large bitmaps.
*
* Initialized flat bitmaps start at all-zeros (all bits unset).
*/
typedef unsigned long fb_group_t;
#define FB_GROUP_BITS (ZU(1) << (LG_SIZEOF_LONG + 3))
#define FB_NGROUPS(nbits) ((nbits) / FB_GROUP_BITS \
+ ((nbits) % FB_GROUP_BITS == 0 ? 0 : 1))
static inline void
fb_init(fb_group_t *fb, size_t nbits) {
size_t ngroups = FB_NGROUPS(nbits);
memset(fb, 0, ngroups * sizeof(fb_group_t));
}
static inline bool
fb_empty(fb_group_t *fb, size_t nbits) {
size_t ngroups = FB_NGROUPS(nbits);
for (size_t i = 0; i < ngroups; i++) {
if (fb[i] != 0) {
return false;
}
}
return true;
}
static inline bool
fb_full(fb_group_t *fb, size_t nbits) {
size_t ngroups = FB_NGROUPS(nbits);
size_t trailing_bits = nbits % FB_GROUP_BITS;
size_t limit = (trailing_bits == 0 ? ngroups : ngroups - 1);
for (size_t i = 0; i < limit; i++) {
if (fb[i] != ~(fb_group_t)0) {
return false;
}
}
if (trailing_bits == 0) {
return true;
}
return fb[ngroups - 1] == ((fb_group_t)1 << trailing_bits) - 1;
}
static inline bool
fb_get(fb_group_t *fb, size_t nbits, size_t bit) {
assert(bit < nbits);
size_t group_ind = bit / FB_GROUP_BITS;
size_t bit_ind = bit % FB_GROUP_BITS;
return (bool)(fb[group_ind] & ((fb_group_t)1 << bit_ind));
}
static inline void
fb_set(fb_group_t *fb, size_t nbits, size_t bit) {
assert(bit < nbits);
size_t group_ind = bit / FB_GROUP_BITS;
size_t bit_ind = bit % FB_GROUP_BITS;
fb[group_ind] |= ((fb_group_t)1 << bit_ind);
}
static inline void
fb_unset(fb_group_t *fb, size_t nbits, size_t bit) {
assert(bit < nbits);
size_t group_ind = bit / FB_GROUP_BITS;
size_t bit_ind = bit % FB_GROUP_BITS;
fb[group_ind] &= ~((fb_group_t)1 << bit_ind);
}
/*
* Some implementation details. This visitation function lets us apply a group
* visitor to each group in the bitmap (potentially modifying it). The mask
* indicates which bits are logically part of the visitation.
*/
typedef void (*fb_group_visitor_t)(void *ctx, fb_group_t *fb, fb_group_t mask);
JEMALLOC_ALWAYS_INLINE void
fb_visit_impl(fb_group_t *fb, size_t nbits, fb_group_visitor_t visit, void *ctx,
size_t start, size_t cnt) {
assert(cnt > 0);
assert(start + cnt <= nbits);
size_t group_ind = start / FB_GROUP_BITS;
size_t start_bit_ind = start % FB_GROUP_BITS;
/*
* The first group is special; it's the only one we don't start writing
* to from bit 0.
*/
size_t first_group_cnt = (start_bit_ind + cnt > FB_GROUP_BITS
? FB_GROUP_BITS - start_bit_ind : cnt);
/*
* We can basically split affected words into:
* - The first group, where we touch only the high bits
* - The last group, where we touch only the low bits
* - The middle, where we set all the bits to the same thing.
* We treat each case individually. The last two could be merged, but
* this can lead to bad codegen for those middle words.
*/
/* First group */
fb_group_t mask = ((~(fb_group_t)0)
>> (FB_GROUP_BITS - first_group_cnt))
<< start_bit_ind;
visit(ctx, &fb[group_ind], mask);
cnt -= first_group_cnt;
group_ind++;
/* Middle groups */
while (cnt > FB_GROUP_BITS) {
visit(ctx, &fb[group_ind], ~(fb_group_t)0);
cnt -= FB_GROUP_BITS;
group_ind++;
}
/* Last group */
if (cnt != 0) {
mask = (~(fb_group_t)0) >> (FB_GROUP_BITS - cnt);
visit(ctx, &fb[group_ind], mask);
}
}
JEMALLOC_ALWAYS_INLINE void
fb_assign_visitor(void *ctx, fb_group_t *fb, fb_group_t mask) {
bool val = *(bool *)ctx;
if (val) {
*fb |= mask;
} else {
*fb &= ~mask;
}
}
/* Sets the cnt bits starting at position start. Must not have a 0 count. */
static inline void
fb_set_range(fb_group_t *fb, size_t nbits, size_t start, size_t cnt) {
bool val = true;
fb_visit_impl(fb, nbits, &fb_assign_visitor, &val, start, cnt);
}
/* Unsets the cnt bits starting at position start. Must not have a 0 count. */
static inline void
fb_unset_range(fb_group_t *fb, size_t nbits, size_t start, size_t cnt) {
bool val = false;
fb_visit_impl(fb, nbits, &fb_assign_visitor, &val, start, cnt);
}
JEMALLOC_ALWAYS_INLINE void
fb_scount_visitor(void *ctx, fb_group_t *fb, fb_group_t mask) {
size_t *scount = (size_t *)ctx;
*scount += popcount_lu(*fb & mask);
}
/* Finds the number of set bit in the of length cnt starting at start. */
JEMALLOC_ALWAYS_INLINE size_t
fb_scount(fb_group_t *fb, size_t nbits, size_t start, size_t cnt) {
size_t scount = 0;
fb_visit_impl(fb, nbits, &fb_scount_visitor, &scount, start, cnt);
return scount;
}
/* Finds the number of unset bit in the of length cnt starting at start. */
JEMALLOC_ALWAYS_INLINE size_t
fb_ucount(fb_group_t *fb, size_t nbits, size_t start, size_t cnt) {
size_t scount = fb_scount(fb, nbits, start, cnt);
return cnt - scount;
}
/*
* An implementation detail; find the first bit at position >= min_bit with the
* value val.
*
* Returns the number of bits in the bitmap if no such bit exists.
*/
JEMALLOC_ALWAYS_INLINE ssize_t
fb_find_impl(fb_group_t *fb, size_t nbits, size_t start, bool val,
bool forward) {
assert(start < nbits);
size_t ngroups = FB_NGROUPS(nbits);
ssize_t group_ind = start / FB_GROUP_BITS;
size_t bit_ind = start % FB_GROUP_BITS;
fb_group_t maybe_invert = (val ? 0 : (fb_group_t)-1);
fb_group_t group = fb[group_ind];
group ^= maybe_invert;
if (forward) {
/* Only keep ones in bits bit_ind and above. */
group &= ~((1LU << bit_ind) - 1);
} else {
/*
* Only keep ones in bits bit_ind and below. You might more
* naturally express this as (1 << (bit_ind + 1)) - 1, but
* that shifts by an invalid amount if bit_ind is one less than
* FB_GROUP_BITS.
*/
group &= ((2LU << bit_ind) - 1);
}
ssize_t group_ind_bound = forward ? (ssize_t)ngroups : -1;
while (group == 0) {
group_ind += forward ? 1 : -1;
if (group_ind == group_ind_bound) {
return forward ? (ssize_t)nbits : (ssize_t)-1;
}
group = fb[group_ind];
group ^= maybe_invert;
}
assert(group != 0);
size_t bit = forward ? ffs_lu(group) : fls_lu(group);
size_t pos = group_ind * FB_GROUP_BITS + bit;
/*
* The high bits of a partially filled last group are zeros, so if we're
* looking for zeros we don't want to report an invalid result.
*/
if (forward && !val && pos > nbits) {
return nbits;
}
return pos;
}
/*
* Find the first set bit in the bitmap with an index >= min_bit. Returns the
* number of bits in the bitmap if no such bit exists.
*/
static inline size_t
fb_ffu(fb_group_t *fb, size_t nbits, size_t min_bit) {
return (size_t)fb_find_impl(fb, nbits, min_bit, /* val */ false,
/* forward */ true);
}
/* The same, but looks for an unset bit. */
static inline size_t
fb_ffs(fb_group_t *fb, size_t nbits, size_t min_bit) {
return (size_t)fb_find_impl(fb, nbits, min_bit, /* val */ true,
/* forward */ true);
}
/*
* Find the last set bit in the bitmap with an index <= max_bit. Returns -1 if
* no such bit exists.
*/
static inline ssize_t
fb_flu(fb_group_t *fb, size_t nbits, size_t max_bit) {
return fb_find_impl(fb, nbits, max_bit, /* val */ false,
/* forward */ false);
}
static inline ssize_t
fb_fls(fb_group_t *fb, size_t nbits, size_t max_bit) {
return fb_find_impl(fb, nbits, max_bit, /* val */ true,
/* forward */ false);
}
/* Returns whether or not we found a range. */
JEMALLOC_ALWAYS_INLINE bool
fb_iter_range_impl(fb_group_t *fb, size_t nbits, size_t start, size_t *r_begin,
size_t *r_len, bool val, bool forward) {
assert(start < nbits);
ssize_t next_range_begin = fb_find_impl(fb, nbits, start, val, forward);
if ((forward && next_range_begin == (ssize_t)nbits)
|| (!forward && next_range_begin == (ssize_t)-1)) {
return false;
}
/* Half open range; the set bits are [begin, end). */
ssize_t next_range_end = fb_find_impl(fb, nbits, next_range_begin, !val,
forward);
if (forward) {
*r_begin = next_range_begin;
*r_len = next_range_end - next_range_begin;
} else {
*r_begin = next_range_end + 1;
*r_len = next_range_begin - next_range_end;
}
return true;
}
/*
* Used to iterate through ranges of set bits.
*
* Tries to find the next contiguous sequence of set bits with a first index >=
* start. If one exists, puts the earliest bit of the range in *r_begin, its
* length in *r_len, and returns true. Otherwise, returns false (without
* touching *r_begin or *r_end).
*/
static inline bool
fb_srange_iter(fb_group_t *fb, size_t nbits, size_t start, size_t *r_begin,
size_t *r_len) {
return fb_iter_range_impl(fb, nbits, start, r_begin, r_len,
/* val */ true, /* forward */ true);
}
/*
* The same as fb_srange_iter, but searches backwards from start rather than
* forwards. (The position returned is still the earliest bit in the range).
*/
static inline bool
fb_srange_riter(fb_group_t *fb, size_t nbits, size_t start, size_t *r_begin,
size_t *r_len) {
return fb_iter_range_impl(fb, nbits, start, r_begin, r_len,
/* val */ true, /* forward */ false);
}
/* Similar to fb_srange_iter, but searches for unset bits. */
static inline bool
fb_urange_iter(fb_group_t *fb, size_t nbits, size_t start, size_t *r_begin,
size_t *r_len) {
return fb_iter_range_impl(fb, nbits, start, r_begin, r_len,
/* val */ false, /* forward */ true);
}
/* Similar to fb_srange_riter, but searches for unset bits. */
static inline bool
fb_urange_riter(fb_group_t *fb, size_t nbits, size_t start, size_t *r_begin,
size_t *r_len) {
return fb_iter_range_impl(fb, nbits, start, r_begin, r_len,
/* val */ false, /* forward */ false);
}
JEMALLOC_ALWAYS_INLINE size_t
fb_range_longest_impl(fb_group_t *fb, size_t nbits, bool val) {
size_t begin = 0;
size_t longest_len = 0;
size_t len = 0;
while (begin < nbits && fb_iter_range_impl(fb, nbits, begin, &begin,
&len, val, /* forward */ true)) {
if (len > longest_len) {
longest_len = len;
}
begin += len;
}
return longest_len;
}
static inline size_t
fb_srange_longest(fb_group_t *fb, size_t nbits) {
return fb_range_longest_impl(fb, nbits, /* val */ true);
}
static inline size_t
fb_urange_longest(fb_group_t *fb, size_t nbits) {
return fb_range_longest_impl(fb, nbits, /* val */ false);
}
/*
* Initializes each bit of dst with the bitwise-AND of the corresponding bits of
* src1 and src2. All bitmaps must be the same size.
*/
static inline void
fb_bit_and(fb_group_t *dst, fb_group_t *src1, fb_group_t *src2, size_t nbits) {
size_t ngroups = FB_NGROUPS(nbits);
for (size_t i = 0; i < ngroups; i++) {
dst[i] = src1[i] & src2[i];
}
}
/* Like fb_bit_and, but with bitwise-OR. */
static inline void
fb_bit_or(fb_group_t *dst, fb_group_t *src1, fb_group_t *src2, size_t nbits) {
size_t ngroups = FB_NGROUPS(nbits);
for (size_t i = 0; i < ngroups; i++) {
dst[i] = src1[i] | src2[i];
}
}
/* Initializes dst bit i to the negation of source bit i. */
static inline void
fb_bit_not(fb_group_t *dst, fb_group_t *src, size_t nbits) {
size_t ngroups = FB_NGROUPS(nbits);
for (size_t i = 0; i < ngroups; i++) {
dst[i] = ~src[i];
}
}
#endif /* JEMALLOC_INTERNAL_FB_H */
#ifndef JEMALLOC_INTERNAL_FXP_H
#define JEMALLOC_INTERNAL_FXP_H
/*
* A simple fixed-point math implementation, supporting only unsigned values
* (with overflow being an error).
*
* It's not in general safe to use floating point in core code, because various
* libc implementations we get linked against can assume that malloc won't touch
* floating point state and call it with an unusual calling convention.
*/
/*
* High 16 bits are the integer part, low 16 are the fractional part. Or
* equivalently, repr == 2**16 * val, where we use "val" to refer to the
* (imaginary) fractional representation of the true value.
*
* We pick a uint32_t here since it's convenient in some places to
* double the representation size (i.e. multiplication and division use
* 64-bit integer types), and a uint64_t is the largest type we're
* certain is available.
*/
typedef uint32_t fxp_t;
#define FXP_INIT_INT(x) ((x) << 16)
#define FXP_INIT_PERCENT(pct) (((pct) << 16) / 100)
/*
* Amount of precision used in parsing and printing numbers. The integer bound
* is simply because the integer part of the number gets 16 bits, and so is
* bounded by 65536.
*
* We use a lot of precision for the fractional part, even though most of it
* gets rounded off; this lets us get exact values for the important special
* case where the denominator is a small power of 2 (for instance,
* 1/512 == 0.001953125 is exactly representable even with only 16 bits of
* fractional precision). We need to left-shift by 16 before dividing by
* 10**precision, so we pick precision to be floor(log(2**48)) = 14.
*/
#define FXP_INTEGER_PART_DIGITS 5
#define FXP_FRACTIONAL_PART_DIGITS 14
/*
* In addition to the integer and fractional parts of the number, we need to
* include a null character and (possibly) a decimal point.
*/
#define FXP_BUF_SIZE (FXP_INTEGER_PART_DIGITS + FXP_FRACTIONAL_PART_DIGITS + 2)
static inline fxp_t
fxp_add(fxp_t a, fxp_t b) {
return a + b;
}
static inline fxp_t
fxp_sub(fxp_t a, fxp_t b) {
assert(a >= b);
return a - b;
}
static inline fxp_t
fxp_mul(fxp_t a, fxp_t b) {
uint64_t unshifted = (uint64_t)a * (uint64_t)b;
/*
* Unshifted is (a.val * 2**16) * (b.val * 2**16)
* == (a.val * b.val) * 2**32, but we want
* (a.val * b.val) * 2 ** 16.
*/
return (uint32_t)(unshifted >> 16);
}
static inline fxp_t
fxp_div(fxp_t a, fxp_t b) {
assert(b != 0);
uint64_t unshifted = ((uint64_t)a << 32) / (uint64_t)b;
/*
* Unshifted is (a.val * 2**16) * (2**32) / (b.val * 2**16)
* == (a.val / b.val) * (2 ** 32), which again corresponds to a right
* shift of 16.
*/
return (uint32_t)(unshifted >> 16);
}
static inline uint32_t
fxp_round_down(fxp_t a) {
return a >> 16;
}
static inline uint32_t
fxp_round_nearest(fxp_t a) {
uint32_t fractional_part = (a & ((1U << 16) - 1));
uint32_t increment = (uint32_t)(fractional_part >= (1U << 15));
return (a >> 16) + increment;
}
/*
* Approximately computes x * frac, without the size limitations that would be
* imposed by converting u to an fxp_t.
*/
static inline size_t
fxp_mul_frac(size_t x_orig, fxp_t frac) {
assert(frac <= (1U << 16));
/*
* Work around an over-enthusiastic warning about type limits below (on
* 32-bit platforms, a size_t is always less than 1ULL << 48).
*/
uint64_t x = (uint64_t)x_orig;
/*
* If we can guarantee no overflow, multiply first before shifting, to
* preserve some precision. Otherwise, shift first and then multiply.
* In the latter case, we only lose the low 16 bits of a 48-bit number,
* so we're still accurate to within 1/2**32.
*/
if (x < (1ULL << 48)) {
return (size_t)((x * frac) >> 16);
} else {
return (size_t)((x >> 16) * (uint64_t)frac);
}
}
/*
* Returns true on error. Otherwise, returns false and updates *ptr to point to
* the first character not parsed (because it wasn't a digit).
*/
bool fxp_parse(fxp_t *a, const char *ptr, char **end);
void fxp_print(fxp_t a, char buf[FXP_BUF_SIZE]);
#endif /* JEMALLOC_INTERNAL_FXP_H */
...@@ -104,8 +104,8 @@ hash_x86_32(const void *key, int len, uint32_t seed) { ...@@ -104,8 +104,8 @@ hash_x86_32(const void *key, int len, uint32_t seed) {
uint32_t k1 = 0; uint32_t k1 = 0;
switch (len & 3) { switch (len & 3) {
case 3: k1 ^= tail[2] << 16; JEMALLOC_FALLTHROUGH case 3: k1 ^= tail[2] << 16; JEMALLOC_FALLTHROUGH;
case 2: k1 ^= tail[1] << 8; JEMALLOC_FALLTHROUGH case 2: k1 ^= tail[1] << 8; JEMALLOC_FALLTHROUGH;
case 1: k1 ^= tail[0]; k1 *= c1; k1 = hash_rotl_32(k1, 15); case 1: k1 ^= tail[0]; k1 *= c1; k1 = hash_rotl_32(k1, 15);
k1 *= c2; h1 ^= k1; k1 *= c2; h1 ^= k1;
} }
...@@ -177,29 +177,29 @@ hash_x86_128(const void *key, const int len, uint32_t seed, ...@@ -177,29 +177,29 @@ hash_x86_128(const void *key, const int len, uint32_t seed,
uint32_t k4 = 0; uint32_t k4 = 0;
switch (len & 15) { switch (len & 15) {
case 15: k4 ^= tail[14] << 16; JEMALLOC_FALLTHROUGH case 15: k4 ^= tail[14] << 16; JEMALLOC_FALLTHROUGH;
case 14: k4 ^= tail[13] << 8; JEMALLOC_FALLTHROUGH case 14: k4 ^= tail[13] << 8; JEMALLOC_FALLTHROUGH;
case 13: k4 ^= tail[12] << 0; case 13: k4 ^= tail[12] << 0;
k4 *= c4; k4 = hash_rotl_32(k4, 18); k4 *= c1; h4 ^= k4; k4 *= c4; k4 = hash_rotl_32(k4, 18); k4 *= c1; h4 ^= k4;
JEMALLOC_FALLTHROUGH JEMALLOC_FALLTHROUGH;
case 12: k3 ^= tail[11] << 24; JEMALLOC_FALLTHROUGH case 12: k3 ^= (uint32_t) tail[11] << 24; JEMALLOC_FALLTHROUGH;
case 11: k3 ^= tail[10] << 16; JEMALLOC_FALLTHROUGH case 11: k3 ^= tail[10] << 16; JEMALLOC_FALLTHROUGH;
case 10: k3 ^= tail[ 9] << 8; JEMALLOC_FALLTHROUGH case 10: k3 ^= tail[ 9] << 8; JEMALLOC_FALLTHROUGH;
case 9: k3 ^= tail[ 8] << 0; case 9: k3 ^= tail[ 8] << 0;
k3 *= c3; k3 = hash_rotl_32(k3, 17); k3 *= c4; h3 ^= k3; k3 *= c3; k3 = hash_rotl_32(k3, 17); k3 *= c4; h3 ^= k3;
JEMALLOC_FALLTHROUGH JEMALLOC_FALLTHROUGH;
case 8: k2 ^= tail[ 7] << 24; JEMALLOC_FALLTHROUGH case 8: k2 ^= (uint32_t) tail[ 7] << 24; JEMALLOC_FALLTHROUGH;
case 7: k2 ^= tail[ 6] << 16; JEMALLOC_FALLTHROUGH case 7: k2 ^= tail[ 6] << 16; JEMALLOC_FALLTHROUGH;
case 6: k2 ^= tail[ 5] << 8; JEMALLOC_FALLTHROUGH case 6: k2 ^= tail[ 5] << 8; JEMALLOC_FALLTHROUGH;
case 5: k2 ^= tail[ 4] << 0; case 5: k2 ^= tail[ 4] << 0;
k2 *= c2; k2 = hash_rotl_32(k2, 16); k2 *= c3; h2 ^= k2; k2 *= c2; k2 = hash_rotl_32(k2, 16); k2 *= c3; h2 ^= k2;
JEMALLOC_FALLTHROUGH JEMALLOC_FALLTHROUGH;
case 4: k1 ^= tail[ 3] << 24; JEMALLOC_FALLTHROUGH case 4: k1 ^= (uint32_t) tail[ 3] << 24; JEMALLOC_FALLTHROUGH;
case 3: k1 ^= tail[ 2] << 16; JEMALLOC_FALLTHROUGH case 3: k1 ^= tail[ 2] << 16; JEMALLOC_FALLTHROUGH;
case 2: k1 ^= tail[ 1] << 8; JEMALLOC_FALLTHROUGH case 2: k1 ^= tail[ 1] << 8; JEMALLOC_FALLTHROUGH;
case 1: k1 ^= tail[ 0] << 0; case 1: k1 ^= tail[ 0] << 0;
k1 *= c1; k1 = hash_rotl_32(k1, 15); k1 *= c2; h1 ^= k1; k1 *= c1; k1 = hash_rotl_32(k1, 15); k1 *= c2; h1 ^= k1;
JEMALLOC_FALLTHROUGH break;
} }
} }
...@@ -261,24 +261,25 @@ hash_x64_128(const void *key, const int len, const uint32_t seed, ...@@ -261,24 +261,25 @@ hash_x64_128(const void *key, const int len, const uint32_t seed,
uint64_t k2 = 0; uint64_t k2 = 0;
switch (len & 15) { switch (len & 15) {
case 15: k2 ^= ((uint64_t)(tail[14])) << 48; JEMALLOC_FALLTHROUGH case 15: k2 ^= ((uint64_t)(tail[14])) << 48; JEMALLOC_FALLTHROUGH;
case 14: k2 ^= ((uint64_t)(tail[13])) << 40; JEMALLOC_FALLTHROUGH case 14: k2 ^= ((uint64_t)(tail[13])) << 40; JEMALLOC_FALLTHROUGH;
case 13: k2 ^= ((uint64_t)(tail[12])) << 32; JEMALLOC_FALLTHROUGH case 13: k2 ^= ((uint64_t)(tail[12])) << 32; JEMALLOC_FALLTHROUGH;
case 12: k2 ^= ((uint64_t)(tail[11])) << 24; JEMALLOC_FALLTHROUGH case 12: k2 ^= ((uint64_t)(tail[11])) << 24; JEMALLOC_FALLTHROUGH;
case 11: k2 ^= ((uint64_t)(tail[10])) << 16; JEMALLOC_FALLTHROUGH case 11: k2 ^= ((uint64_t)(tail[10])) << 16; JEMALLOC_FALLTHROUGH;
case 10: k2 ^= ((uint64_t)(tail[ 9])) << 8; JEMALLOC_FALLTHROUGH case 10: k2 ^= ((uint64_t)(tail[ 9])) << 8; JEMALLOC_FALLTHROUGH;
case 9: k2 ^= ((uint64_t)(tail[ 8])) << 0; case 9: k2 ^= ((uint64_t)(tail[ 8])) << 0;
k2 *= c2; k2 = hash_rotl_64(k2, 33); k2 *= c1; h2 ^= k2; k2 *= c2; k2 = hash_rotl_64(k2, 33); k2 *= c1; h2 ^= k2;
JEMALLOC_FALLTHROUGH JEMALLOC_FALLTHROUGH;
case 8: k1 ^= ((uint64_t)(tail[ 7])) << 56; JEMALLOC_FALLTHROUGH case 8: k1 ^= ((uint64_t)(tail[ 7])) << 56; JEMALLOC_FALLTHROUGH;
case 7: k1 ^= ((uint64_t)(tail[ 6])) << 48; JEMALLOC_FALLTHROUGH case 7: k1 ^= ((uint64_t)(tail[ 6])) << 48; JEMALLOC_FALLTHROUGH;
case 6: k1 ^= ((uint64_t)(tail[ 5])) << 40; JEMALLOC_FALLTHROUGH case 6: k1 ^= ((uint64_t)(tail[ 5])) << 40; JEMALLOC_FALLTHROUGH;
case 5: k1 ^= ((uint64_t)(tail[ 4])) << 32; JEMALLOC_FALLTHROUGH case 5: k1 ^= ((uint64_t)(tail[ 4])) << 32; JEMALLOC_FALLTHROUGH;
case 4: k1 ^= ((uint64_t)(tail[ 3])) << 24; JEMALLOC_FALLTHROUGH case 4: k1 ^= ((uint64_t)(tail[ 3])) << 24; JEMALLOC_FALLTHROUGH;
case 3: k1 ^= ((uint64_t)(tail[ 2])) << 16; JEMALLOC_FALLTHROUGH case 3: k1 ^= ((uint64_t)(tail[ 2])) << 16; JEMALLOC_FALLTHROUGH;
case 2: k1 ^= ((uint64_t)(tail[ 1])) << 8; JEMALLOC_FALLTHROUGH case 2: k1 ^= ((uint64_t)(tail[ 1])) << 8; JEMALLOC_FALLTHROUGH;
case 1: k1 ^= ((uint64_t)(tail[ 0])) << 0; case 1: k1 ^= ((uint64_t)(tail[ 0])) << 0;
k1 *= c1; k1 = hash_rotl_64(k1, 31); k1 *= c2; h1 ^= k1; k1 *= c1; k1 = hash_rotl_64(k1, 31); k1 *= c2; h1 ^= k1;
break;
} }
} }
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment