)]}' { "log": [ { "commit": "5d427bcb08d74112701cc6c6bf46d7344ea07cc7", "tree": "9e9091c92ff108909d0ff2ff7091a01a075cbcf7", "parents": [ "e9bd18ac0c1413473211bbd427c12275b92c9f0e" ], "author": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Wed Nov 07 14:48:29 2018 -0800" }, "committer": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Mon Nov 12 13:42:28 2018 -0800" }, "message": "bionic: Add shadow call stack guard region field to pthread_internal_t.\n\nLanding this change separately to the change that implements SCS\nbecause it needs to land at the same time as an internal change. This\nwill simplify the situation in case SCS needs to be reverted again.\n\nChange-Id: Ibe18750829665b6dcf6e36628a5e5bbdd1a0dd4f\nMerged-In: Ibe18750829665b6dcf6e36628a5e5bbdd1a0dd4f\n" }, { "commit": "f90687c4ec2eba944fa697eff2ccaf9e3b8ee480", "tree": "c4ef71eeb0bce234bd0a4ad8dac35e8778de7b75", "parents": [ "0c39bd468d3aa19b3e27937b36b101cc4aa65dab" ], "author": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Tue Nov 06 14:47:27 2018 -0800" }, "committer": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Thu Nov 08 21:24:31 2018 +0000" }, "message": "Increase the size of the signal stack on 64-bit.\n\nThe signal stack is sufficiently large for a single invocation of the\nsignal handler, but in cases where the signal handler needs to recurse,\n(e.g. if our address space is limited by RLIMIT_AS), it\u0027s too small for\nus to get to the part where we recognize that we\u0027ve recursed and bail\nout.\n\nBug: http://b/118772392\nTest: /data/nativetest64/debuggerd_test/debuggerd_test64 --gtest_filter\u003dCrasherTest.seccomp_crash_oom\nChange-Id: Ic7a2cf8b01b3f7ea7f4a2318a3ec22a0c3649da6\n" }, { "commit": "6f3a56bb18628243b6dbe470222e56bb56ed10ae", "tree": "53a7886e2e1f2c7766986c5eafe3f631fe6027ce", "parents": [ "5c7bece8ca071118cceb42306ca568df495d09c7", "7b13dd918e6e2c1638e6b137f27c80979a4ecd3f" ], "author": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Thu Nov 01 01:32:00 2018 +0000" }, "committer": { "name": "Gerrit Code Review", "email": "noreply-gerritcodereview@google.com", "time": "Thu Nov 01 01:32:00 2018 +0000" }, "message": "Merge \"Revert \"bionic: Allocate a shadow call stack for each thread.\"\"" }, { "commit": "25c34e0feba2d0c49971016117a5af4fb8d32c1e", "tree": "46b26664f7fd968f0009981b5f25878239227d41", "parents": [ "f284a475441f612a0b48067a8bc7a59c72c4b47a" ], "author": { "name": "Ryan Prichard", "email": "rprichard@google.com", "time": "Mon Oct 29 16:31:51 2018 -0700" }, "committer": { "name": "Ryan Prichard", "email": "rprichard@google.com", "time": "Tue Oct 30 12:30:21 2018 -0700" }, "message": "Reverse two pthread_internal_t fields\n\nThe golang runtime currently expects to find the pthread key data after\nthe tls slots.\n\nBug: http://b/78026329\nBug: http://b/118381796\nTest: run a golang-based app, bionic unit tests\nChange-Id: Idc777d809b803093e1c81d9a2ce4eafcc7d61f8d\nMerged-In: Idc777d809b803093e1c81d9a2ce4eafcc7d61f8d\n(cherry picked from commit a2c30723dac6b570247ac1bc7190850616f4db93)\n" }, { "commit": "7b13dd918e6e2c1638e6b137f27c80979a4ecd3f", "tree": "b7f8343a6fc3767bd968b68ba8b6be74466d027a", "parents": [ "da1bc79f937225b1a048d9e5a03eca81680a17fd" ], "author": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Mon Oct 29 21:48:38 2018 +0000" }, "committer": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Mon Oct 29 21:48:59 2018 +0000" }, "message": "Revert \"bionic: Allocate a shadow call stack for each thread.\"\n\nThis reverts commit da1bc79f937225b1a048d9e5a03eca81680a17fd.\n\nReason for revert: Caused OOM in media process\n\nBug: 112907825\nBug: 118593766\nChange-Id: I545663871d75889b209b9fd2131cdaa97166478f\n" }, { "commit": "f284a475441f612a0b48067a8bc7a59c72c4b47a", "tree": "47d3953dcd5c84de7b57e01fae11d3c0cab606dd", "parents": [ "debcd02d184822df57e7dc69066933dd44092235", "da1bc79f937225b1a048d9e5a03eca81680a17fd" ], "author": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Mon Oct 29 21:05:25 2018 +0000" }, "committer": { "name": "Gerrit Code Review", "email": "noreply-gerritcodereview@google.com", "time": "Mon Oct 29 21:05:25 2018 +0000" }, "message": "Merge \"bionic: Allocate a shadow call stack for each thread.\"" }, { "commit": "da1bc79f937225b1a048d9e5a03eca81680a17fd", "tree": "0334b394fff71fc1263d82515dde5c8c16a5dda8", "parents": [ "c89a3971e94c30611c885e4c440eae2fd9937948" ], "author": { "name": "Peter Collingbourne", "email": "pcc@google.com", "time": "Thu Sep 06 22:20:44 2018 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Oct 29 19:55:18 2018 +0000" }, "message": "bionic: Allocate a shadow call stack for each thread.\n\nBug: 112907825\nChange-Id: I7c1479a0cd68696739bf6aa5e0700ba4f2a137ec\nMerged-In: I7c1479a0cd68696739bf6aa5e0700ba4f2a137ec\n" }, { "commit": "29d6dbc906f51f127700fd9e2d1e1170456f2382", "tree": "09bdf2be3f29c92de8ac17f8ee27691d26356640", "parents": [ "a2af8bea8c6c4b6ce6883bf92cbce922e0d9f458" ], "author": { "name": "Ryan Prichard", "email": "rprichard@google.com", "time": "Fri Oct 19 20:35:23 2018 -0700" }, "committer": { "name": "Ryan Prichard", "email": "rprichard@google.com", "time": "Thu Oct 25 23:31:17 2018 +0000" }, "message": "Move Bionic slots to the end of pthread_internal_t\n\nWith ELF TLS, the static linker assumes that an executable\u0027s TLS segment\nis at a known offset from the thread pointer (i.e. __get_tls()). The\nsegment can be located prior to the TP (variant 2, x86[_64], sparc) or\nafter it (variant 1, arm{32,64}, ppc, mips, ia64, riscv).\n\nWe can\u0027t make our pthread_internal_t exactly follow the ordinary arm64\nABI (at least) because TP[5] is used for clang\u0027s -fstack-protector on\nAndroid. Instead, reserve extra space after the TP (up to 16 words), which\nwill be followed by the executable\u0027s TLS segment.\n\nBug: http://b/78026329\nTest: boot device, bionic unit tests\nChange-Id: I0f3b270b793f9872ba0effeac03f4dec364438d6\nMerged-In: I0f3b270b793f9872ba0effeac03f4dec364438d6\n(cherry picked from commit f397317e96928ef24b980d5c73d08829c0e70cd4)\n" }, { "commit": "13e8dcbdc9f0b3d95cac1a234730061c9b1ae6f4", "tree": "0c17f125a779e03f9911ce092e16b0ada1aa2920", "parents": [ "a02d631163860c795cac7a0428ad8abb9ed3b747" ], "author": { "name": "Evgenii Stepanov", "email": "eugenis@google.com", "time": "Wed Sep 19 16:29:12 2018 -0700" }, "committer": { "name": "Evgenii Stepanov", "email": "eugenis@google.com", "time": "Thu Sep 20 12:20:12 2018 -0700" }, "message": "[hwasan] Tweak process and thread initialization.\n\nMake sure that TLS_SLOT_TSAN is always available and correctly set up in\nHWASan-instrumented functions by setting up the tls register and running hwasan\ninitialization (__hwasan_init in the main thread and __hwasan_thread_enter in\nsecondary) early enough.\n\nThis is needed to accomodate a change in HWASan: https://reviews.llvm.org/D52249\n\nBug: 112438058\nTest: boot with SANITIZE_TARGET\u003dhwaddress, run bionic-tests\nChange-Id: Icd909a4ea0da6c6c1095522bcc28debef5f2c63d\n" }, { "commit": "be551f596f763e147bb29ffc26bc360d910f90fb", "tree": "f0a7a9aa02e1b73e840f62262125d8064a9d4aae", "parents": [ "3fd45bba4857fdbf320b6e89d2ae0569d9463bf5" ], "author": { "name": "Evgenii Stepanov", "email": "eugenis@google.com", "time": "Mon Aug 13 16:46:15 2018 -0700" }, "committer": { "name": "Evgenii Stepanov", "email": "eugenis@google.com", "time": "Tue Aug 21 00:15:47 2018 +0000" }, "message": "HWASan support in bionic.\n\n* Allow sanitization of libc (excluding existing global sanitizers)\n and disallow sanitization of linker. The latter has not been\n necessary before because HWASan is the first sanitizer to support\n static binaries (with the exception of CFI, which is not used\n globally).\n* Static binary startup: initialize HWASan shadow very early so that\n almost entire libc can be sanitized. The rest of initialization is\n done in a global constructor; until that is done sanitized code can\n run but can\u0027t report errors (will simply crash with SIGTRAP).\n* Switch malloc_common from je_* to __sanitizer_*.\n* Call hwasan functions when entering and leaving threads. We can not\n intercept pthread_create when libc depends on libclang_rt.hwasan.\n An alternative to this would be a callback interface like requested\n here:\n https://sourceware.org/glibc/wiki/ThreadPropertiesAPI\n\nAll of the above is behind a compile-time check\n__has_feature(hwaddress_sanitizer). This means that HWASan actually\nrequires libc to be instrumented, and would not work otherwise. It\u0027s\nan implementation choice that greatly reduces complexity of the tool.\nInstrumented libc also guarantees that hwasan is present and\ninitialized in every process, which allows piecemeal sanitization\n(i.e. library w/o main executable, or even individual static\nlibraries), unlike ASan.\n\nChange-Id: If44c46b79b15049d1745ba46ec910ae4f355d19c\n" }, { "commit": "cbc80ba9d839675a0c4891e2ab33f39ba51b04b2", "tree": "0b24e6123c94bd3e61d94fd6eca6a3b9cc9b074e", "parents": [ "3374d0cabb973ff2b73ad8cd909389bdeed18658" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 13 14:26:29 2018 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 13 14:27:17 2018 -0800" }, "message": "Switch the rest of our internal headers to #pragma once.\n\nWe\u0027ve been using #pragma once for new internal files, but let\u0027s be more bold.\n\nBug: N/A\nTest: builds\nChange-Id: I7e2ee2730043bd884f9571cdbd8b524043030c07\n" }, { "commit": "38f01e05ef7f7ef18a43339436fba645d123b838", "tree": "e6dbb0977a34d052056e935d86bd3cdcc9473188", "parents": [ "dfece7a3ee1b8a11abea3f55bf44df21ebc4bc79" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Oct 27 15:28:54 2017 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Oct 27 15:40:15 2017 -0700" }, "message": "Preserve historical pthread_create scheduler behavior better.\n\nAt the cost of two flag bits for what POSIX thinks should be a boolean\nchoice, plus somewhat confusing behavior from pthread_attr_getinheritsched\ndepending on when you call it/what specific scheduler attributes you\u0027ve\nset in the pthread_attr_t, we can emulate the old behavior exactly and\nprevent annoying SELinux denial spam caused by calls to sched_setscheduler.\n\nBug: http://b/68391226\nTest: adb logcat on boot contains no sys_nice avc denials\nChange-Id: I4f759c2c4fd1d80cceb0912d7da09d35902e2e5e\n" }, { "commit": "8aecba7aa6b7f7b92f69c0d3febef59fdb135f87", "tree": "fd3d9743a0e282dcbb2a8b2b51906f7da3a84468", "parents": [ "435e6384de8f9e35b8878b1ccda5bb5686c15207" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Oct 17 15:34:41 2017 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Oct 25 14:28:42 2017 -0700" }, "message": "Implement pthread_attr_getinheritsched/pthread_attr_setinheritsched.\n\nHistorically, Android defaulted to EXPLICIT but with a special case\nbecause SCHED_NORMAL/priority 0 was awkward. Because the code couldn\u0027t\nactually tell whether SCHED_NORMAL/priority 0 was a genuine attempt to\nexplicitly set those attributes (because the parent thread is SCHED_FIFO,\nsay) or just because the pthread_attr_t was left at its defaults.\n\nNow we support INHERIT, we could call sched_getscheduler to see whether\nwe actually need to call sched_setscheduler, but since the major cost\nis the fixed syscall overhead, we may as well just conservatively\ncall sched_setscheduler and let the kernel decide whether it\u0027s a\nno-op. (Especially because we\u0027d then have to add both sched_getscheduler\nand sched_setscheduler to any seccomp filter.)\n\nPlatform code (or app code that only needs to support \u003e\u003d P) can actually\nadd a call to pthread_attr_setinheritsched to say that they just want\nto inherit (if they know that none of their threads actually mess with\nscheduler attributes at all), which will save them a sched_setscheduler\ncall except in the doubly-special case of SCHED_RESET_ON_FORK (which we\ndo handle).\n\nAn alternative would be \"make pthread_attr_setschedparams and\npthread_attr_setschedprio set EXPLICIT and change the platform default\nto INHERIT\", but even though I can only think of weird pathological\nexamples where anyone would notice that change, that behavior -- of\npthread_attr_setschedparams/pthread_attr_setschedprio overriding an\nearlier call to pthread_attr_setinheritsched -- isn\u0027t allowed by POSIX\n(whereas defaulting to EXPLICIT is).\n\nIf we have a lot of trouble with this change in the app compatibility\ntesting phase, though, we\u0027ll want to reconsider this decision!\n\n -*-\n\nThis change also removes a comment about setting the scheduler attributes\nin main_thread because we\u0027d have to actually keep them up to date,\nand it\u0027s not clear that doing so would be worth the trouble.\n\nAlso make async_safe_format_log preserve errno so we don\u0027t have to be\nso careful around it.\n\nBug: http://b/67471710\nTest: ran tests\nChange-Id: Idd026c4ce78a536656adcb57aa2e7b2c616eeddf\n" }, { "commit": "53dc9dd70155fd75af744cbebecc563658c69818", "tree": "2c115be9ba0a403a6a42d9e880746275313dbd79", "parents": [ "cce6ada00ad055ac52791526e44b1f223bec8ce4" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Sep 19 14:02:50 2017 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Sep 19 14:02:50 2017 -0700" }, "message": "Improve pthread_create failure handling.\n\nReturn EAGAIN rather than aborting if we fail to set up the TLS for a new\nthread.\n\nAdd a test that uses all the VMAs so we can properly test these edge cases.\n\nAdd an explicit test for pthread_attr_setdetachstate, which we use in the\nprevious test, but other than that has no tests.\n\nRemove support for ro.logd.timestamp/persist.logd.timestamp, which doesn\u0027t\nseem to be used, and which prevents us from logging failures in cases where\nmmap fails (because we need to mmap in the system property implementation).\n\nBug: http://b/65608572\nTest: ran tests\nChange-Id: I9009f06546e1c2cc55eff996d08b55eff3482343\n" }, { "commit": "d6c678ca90d6f7843a9186515fa38d9eec467ff6", "tree": "bb6316d48f1ab523c0520bcba316b2ba232dfa34", "parents": [ "0d5d0746e8b88dab66038d6150cb92b62467aac1" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Jun 27 17:01:57 2017 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Sep 18 16:09:43 2017 -0700" }, "message": "Support larger guard regions.\n\nThis also fixes a long-standing bug where the guard region would be taken\nout of the stack itself, rather than being -- as POSIX demands -- additional\nspace after the stack. Historically a 128KiB stack with a 256KiB guard would\nhave given you an immediate crash.\n\nBug: http://b/38413813\nTest: builds, boots\nChange-Id: Idd12a3899be1d92fea3d3e0fa6882ca2216bd79c\n" }, { "commit": "415daa8cca875d348c003a95cf7c44c9231eae75", "tree": "0846b1476c61bb7d870fc8e031099d634c0c2e3c", "parents": [ "e429bf2f696282537efb577feb7cb9fd534e7ace" ], "author": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Mon Mar 06 17:45:33 2017 -0800" }, "committer": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Wed Mar 08 16:43:59 2017 -0800" }, "message": "Increase signal stack size on 32-bit to 16kB.\n\nsnprintf to a buffer of length PATH_MAX consumes about 7kB of stack.\n\nBug: http://b/35858739\nTest: bionic-unit-tests --gtest_filter\u003d\"*big_enough*\"\nChange-Id: I34a7f42c1fd2582ca0d0a9b7e7a5290bc1cc19b1\n" }, { "commit": "5e2285d3ccdbb64a49ad2e5e521f50c897a3954d", "tree": "cd56c251f1710652fad3b84121248a12071b42e0", "parents": [ "c9a840ac76c7cdbe6028ac91cdb8eb6698f0a854" ], "author": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Wed Feb 22 12:19:05 2017 -0800" }, "committer": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Wed Feb 22 16:05:03 2017 -0800" }, "message": "Allocate thread local buffers in __init_tls.\n\nThread local buffers were using pthread_setspecific for storage with\nlazy initialization. pthread_setspecific shares TLS slots between the\nlinker and libc.so, so thread local buffers being initialized in a\ndifferent order between libc.so and the linker meant that bad things\nwould happen (manifesting as snprintf not working because the\nlocale was mangled)\n\nBug: http://b/20464031\nTest: /data/nativetest64/bionic-unit-tests/bionic-unit-tests\n everything passes\nTest: /data/nativetest/bionic-unit-tests/bionic-unit-tests\n thread_local tests are failing both before and after (KUSER_HELPERS?)\nTest: /data/nativetest64/bionic-unit-tests-static/bionic-unit-tests-static\n no additional failures\nChange-Id: I9f445a77c6e86979f3fa49c4a5feecf6ec2b0c3f\n" }, { "commit": "7484c21c4c352a2200d94939fabc10d1bd3f0723", "tree": "fe089ec40970a76f11107534632b1fc5789660ed", "parents": [ "b0e8c565a622b5519e03d4416b0b5b1a5f20d7f5" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Feb 02 02:41:38 2017 +0000" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Feb 02 02:41:38 2017 +0000" }, "message": "Revert \"Remove the global thread list.\"\n\nThis reverts commit b0e8c565a622b5519e03d4416b0b5b1a5f20d7f5.\n\nBreaks swiftshader (http:/b/34883464).\n\nChange-Id: I7b21193ba8a78f07d7ac65e41d0fe8516940a83b\n" }, { "commit": "b0e8c565a622b5519e03d4416b0b5b1a5f20d7f5", "tree": "8362f2754276b72e79a652885be7379127ca6e07", "parents": [ "fb07c36bc061db4ca5d8348ff6bc1e60b6c53191" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jan 04 14:12:54 2017 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Sat Jan 07 14:16:46 2017 -0800" }, "message": "Remove the global thread list.\n\nAnother release, another attempt to fix this bug.\n\nThis change affects pthread_detach, pthread_getcpuclockid,\npthread_getschedparam/pthread_setschedparam, pthread_join, and pthread_kill:\ninstead of returning ESRCH when passed an invalid pthread_t, they\u0027ll now SEGV.\n\nNote that this doesn\u0027t change behavior as much as you might think: the old\nlookup only held the global thread list lock for the duration of the lookup,\nso there was still a race between that and the dereference in the caller,\ngiven that callers actually need the tid to pass to some syscall or other,\nand sometimes update fields in the pthread_internal_t struct too.\n\nWe can\u0027t check thread-\u003etid against 0 to see whether a pthread_t is still\nvalid because a dead thread gets its thread struct unmapped along with its\nstack, so the dereference isn\u0027t safe.\n\nTaking the affected functions one by one:\n\n* pthread_getcpuclockid and pthread_getschedparam/pthread_setschedparam\n should be fine. Unsafe calls to those seem highly unlikely.\n\n* Unsafe pthread_detach callers probably want to switch to\n pthread_attr_setdetachstate instead, or using pthread_detach(pthread_self())\n from the new thread\u0027s start routine rather than doing the detach in the\n parent.\n\n* pthread_join calls should be safe anyway, because a joinable thread won\u0027t\n actually exit and unmap until it\u0027s joined. If you\u0027re joining an\n unjoinable thread, the fix is to stop marking it detached. If you\u0027re\n joining an already-joined thread, you need to rethink your design.\n\n* Unsafe pthread_kill calls aren\u0027t portably fixable. (And are obviously\n inherently non-portable as-is.) The best alternative on Android is to\n use pthread_gettid_np at some point that you know the thread to be alive,\n and then call kill/tgkill directly. That\u0027s still not completely safe\n because if you\u0027re too late, the tid may have been reused, but then your\n code is inherently unsafe anyway.\n\nIf we find too much code is still broken, we can come back and disable\nthe global thread list lookups for anything targeting \u003e\u003d O and then have\nanother go at really removing this in P...\n\nBug: http://b/19636317\nTest: N6P boots, bionic tests pass\nChange-Id: Ia92641212f509344b99ee2a9bfab5383147fcba6\n" }, { "commit": "b6453c52ac55f85d7f88f04db6e320825cea9bf7", "tree": "b1ecbacb334303b5e0e0c2773700977eeb97b574", "parents": [ "1089afb744de588d841ffdeed158dbcd113a8e02" ], "author": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Wed Jun 29 16:47:53 2016 -0700" }, "committer": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Thu Jun 30 12:58:32 2016 -0700" }, "message": "Only initialize the global stack protector once.\n\nBefore, dynamic executables would initialize the global stack protector\ntwice, once for the linker, and once for the executable. This worked\nbecause the result was the same for both initializations, because it\nused getauxval(AT_RANDOM), which won\u0027t be the case once arc4random gets\nused for it.\n\nBug: http://b/29622562\nChange-Id: I7718b1ba8ee8fac7127ab2360cb1088e510fef5c\nTest: ran the stack protector tests on angler (32/64bit, static/dynamic)\n" }, { "commit": "1089afb744de588d841ffdeed158dbcd113a8e02", "tree": "7ce78b2a4780ccf8be0841a900354e2613149929", "parents": [ "1159c196db79993b153114bcc4363f7e0f831773" ], "author": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Wed Jun 29 16:19:25 2016 -0700" }, "committer": { "name": "Josh Gao", "email": "jmgao@google.com", "time": "Thu Jun 30 12:58:32 2016 -0700" }, "message": "Make getpid work before TLS has been initialized.\n\nBug: http://b/29622562\nChange-Id: I648adc35c04604a7e8bc649c425f07a723e96d3a\nTest: code dependent on this change no longer crashes\n" }, { "commit": "ff624c2c174b3f19e8c6e0c8f782cdd0caa26f74", "tree": "38fa4a1037feb3a24fcdbbd968018ef10416cba8", "parents": [ "eaf50be332f015bd90e45d4f9d111b24cd230f00" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Wed Mar 30 17:48:50 2016 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Mon Apr 04 16:19:39 2016 -0700" }, "message": "Format code to calculate thread stack and signal stack.\n\nThe code to calculate thread stack and signal stack looks weird:\nthe thread stack size and signal stack size are related with\neach other on 32-bit mode, but not on 64-bit mode. So change the\ncode to make the logic more resonable. This doesn\u0027t change anything\nas we have defined SIGSTKSZ to 16K on arm64.\n\nBug: 28005110\n\nChange-Id: I04d2488cfb96ee7e2d894d062c66cef950fec418\n" }, { "commit": "42d949ff9d2956e25f84e537f43a8f93ecb37baf", "tree": "66f1215029c3777d59393fbbbaa55ba5aed5872a", "parents": [ "988e71b2b22117b0f7e994cbe764066630494125" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jan 06 19:51:43 2016 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jan 06 20:06:08 2016 -0800" }, "message": "Defend against -fstack-protector in libc startup.\n\nExactly which functions get a stack protector is up to the compiler, so\nlet\u0027s separate the code that sets up the environment stack protection\nrequires and explicitly build it with -fno-stack-protector.\n\nBug: http://b/26276517\nChange-Id: I8719e23ead1f1e81715c32c1335da868f68369b5\n" }, { "commit": "952e9eb0866bc2061b671e8d6c90cae275eda6bd", "tree": "1a6fd63119783ddb62d9faaec1e57489b707fc18", "parents": [ "8bd27182c505cdc8031ab23fea5d1dccb6b7405a" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Nov 24 17:24:06 2015 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Nov 24 17:24:06 2015 -0800" }, "message": "Don\u0027t use __thread in __cxa_thread_finalize().\n\nCurrently we use __thread variable to store thread_local_dtors,\nwhich makes tsan test fork_atexit.cc hang. The problem is as below:\nThe main thread creates a worker thread, the worker thread calls\npthread_exit() -\u003e __cxa_thread_finalize() -\u003e __emutls_get_address()\n-\u003e pthread_once(emutls_init) -\u003e emutls_init().\nThen the main thread calls fork(), the child process cals\nexit() -\u003e __cxa_thread_finalize() -\u003e __emutls_get_address()\n-\u003e pthread_once(emutls_init).\nSo the child process is waiting for pthread_once(emutls_init)\nto finish which will never occur.\n\nIt might be the test\u0027s fault because POSIX standard says if a\nmulti-threaded process calls fork(), the new process may only\nexecute async-signal-safe operations until exec functions are\ncalled. And exit() is not async-signal-safe. But we can make\nbionic more reliable by not using __thread in\n__cxa_thread_finalize().\n\nBug: 25392375\nChange-Id: Ife403dd7379dad8ddf1859c348c1c0adea07afb3\n" }, { "commit": "d26e780df66b9add4cf7e7ebb2f6c6749d1c5050", "tree": "117eec6a39195e7edd7ead370513436698734abc", "parents": [ "5edf077c5b92f0db212e45ed5402339a6e4c7334" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Oct 22 20:07:56 2015 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Oct 22 20:14:33 2015 -0700" }, "message": "Use bionic lock in pthread_internal_t.\n\nIt removes calling to pthread_mutex_lock() at the beginning of new\nthread, which helps to support thread sanitizer.\n\nChange-Id: Ia3601c476de7976a9177b792bd74bb200cee0e13\n" }, { "commit": "33ac04a215a70d2f39f1ddec1bba5a42c0fd8bee", "tree": "481db0a7986cd82b887f82d53d759a6afee88f22", "parents": [ "1399759118f50071d499f88b18828e36aa2bad00" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Sep 22 11:16:15 2015 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Sep 22 11:18:26 2015 -0700" }, "message": "Increase alternative signal stack size on 64-bit devices.\n\nBug: 23041777\nBug: 24187462\nChange-Id: I7d84c0cc775a74753a3e8e101169c0fb5dbf7437\n" }, { "commit": "ef115003012f61cf5539fdfeb201b98e4a92f610", "tree": "d546925be11de056c9c2611fb75d91273cd10303", "parents": [ "8225ad63fac7c5cc298884c101a344959b87dc39" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Mon Mar 30 20:03:57 2015 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Mon Mar 30 20:51:39 2015 -0700" }, "message": "Revert \"Revert \"add guard pages to the internal signal stacks\"\"\n\nThis reverts commit a3125fd1396a09a7fc4872dc4653f342150a3deb.\nAnd Fix the prctl() problem that cause system crash.\n\nChange-Id: Icc8d12d848cfba881a7984ca2827fd81be41f9fd\n" }, { "commit": "a3125fd1396a09a7fc4872dc4653f342150a3deb", "tree": "5042f453214b2d176c601710c87350c30c7f3944", "parents": [ "595752f623ae88f7e4193a6e531a0805f1c6c4dc" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Mar 31 02:42:39 2015 +0000" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Mar 31 02:42:39 2015 +0000" }, "message": "Revert \"add guard pages to the internal signal stacks\"\n\nThis reverts commit 595752f623ae88f7e4193a6e531a0805f1c6c4dc.\n\nChange-Id: Iefa66e9049ca0424e53cd5fc320d161b93556dcb\n" }, { "commit": "595752f623ae88f7e4193a6e531a0805f1c6c4dc", "tree": "2a8d25e25b042f04146725c7a287619e2892940c", "parents": [ "203082b8217cd853eee510ad371b382701960610" ], "author": { "name": "Daniel Micay", "email": "danielmicay@gmail.com", "time": "Sun Mar 29 00:03:55 2015 -0400" }, "committer": { "name": "Daniel Micay", "email": "danielmicay@gmail.com", "time": "Mon Mar 30 17:13:20 2015 -0400" }, "message": "add guard pages to the internal signal stacks\n\nSignal handlers tend to be lean, but can still overflow the (tiny)\nstack.\n\nChange-Id: Ia21c6453d92a9f8d1536ad01ff26a1a84c05f8fb\n" }, { "commit": "93d44ff2a6e0febe18d42d5690e85023545318b4", "tree": "0145c42a21091f0a160129bd2978e9a7baefe543", "parents": [ "32b039e46eee4d40b9979eab873a957e86fbf969", "673b15e4ee2c6d99b150aedddc0f389e29f98e1b" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Mar 24 02:39:50 2015 +0000" }, "committer": { "name": "Gerrit Code Review", "email": "noreply-gerritcodereview@google.com", "time": "Tue Mar 24 02:39:51 2015 +0000" }, "message": "Merge \"Let g_thread_list_lock only protect g_thread_list.\"" }, { "commit": "673b15e4ee2c6d99b150aedddc0f389e29f98e1b", "tree": "edc101b54c689960bdf1262e6378b21082d3e277", "parents": [ "4149dc944bbe3c5f1c14fc998106975e68bdd6bf" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Mar 19 14:19:19 2015 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Mon Mar 23 19:03:49 2015 -0700" }, "message": "Let g_thread_list_lock only protect g_thread_list.\n\nAs glibc/netbsd don\u0027t protect access to thread struct members by a global\nlock, we don\u0027t want to do it either. This change reduces the\nresponsibility of g_thread_list_lock to only protect g_thread_list.\n\nBug: 19636317\nChange-Id: I897890710653dac165d8fa4452c7ecf74abdbf2b\n" }, { "commit": "a2db50d5d7fa67b297eddd1c0549f08ea4b6a950", "tree": "e3b0bb4b01b2b7caf8f5f0ad5be05517592915fb", "parents": [ "4149dc944bbe3c5f1c14fc998106975e68bdd6bf" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Mar 20 10:58:04 2015 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Mar 20 14:41:52 2015 -0700" }, "message": "Fix alignment error for pthread_internal_t/pthread stack.\n\naligned attribute can only control compiler\u0027s behavior, but we\nare manually allocating pthread_internal_t. So we need to make\nsure of alignment manually.\n\nChange-Id: Iea4c46eadf10dfd15dc955c5f41cf6063cfd8536\n" }, { "commit": "aec2bb5ec67637ec4b5374287e9bc0d572632054", "tree": "81a349d5c4f2a4962201fa9aea395014bdc89867", "parents": [ "2e49508187254903943b83df7bb241d5fb6a8593" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Mar 19 16:28:19 2015 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Mar 19 23:35:23 2015 +0000" }, "message": "Fix two errors in pthread_detach.cpp.\n\nThe errors are introduced in \"Make pthread join_state not protected by g_thread_list_lock\".\n\nBug: 19636317\nChange-Id: I58ae9711da94bfbac809abfd81311eeb70301a4b\n" }, { "commit": "2f836d4989845c0c82a1e4f99206fb0ff0d137a2", "tree": "7190db5cf1a8d85e40aa664b7ac50aa120ace278", "parents": [ "c0eaa732844a8397d40a7ec82641f709e4b47a61" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Wed Mar 18 14:14:02 2015 -0700" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Wed Mar 18 19:35:34 2015 -0700" }, "message": "Make __get_thread inlined.\n\nBug: 19825434\n\nChange-Id: Ifb672a45a5776b83625a25654ed0d6f7fc368ae3\n" }, { "commit": "58cf31b50699ed9f523de38c8e943f3bbd1ced9e", "tree": "258ba808074bd047a94df452a4e465db66136797", "parents": [ "e86a86f9f24df7028d2596c69ff008cf88e039e4" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Mar 06 17:23:53 2015 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Mar 12 21:39:49 2015 -0700" }, "message": "Make pthread join_state not protected by g_thread_list_lock.\n\n1. Move the representation of thread join_state from pthread.attr.flag\n to pthread.join_state. This clarifies thread state change.\n2. Use atomic operations for pthread.join_state. So we don\u0027t need to\n protect it by g_thread_list_lock. g_thread_list_lock will be reduced\n to only protect g_thread_list or even removed in further changes.\n\nBug: 19636317\nChange-Id: I31fb143a7c69508c7287307dd3b0776993ec0f43\n" }, { "commit": "9d0c79304d2413028aa5c213f7a567f00feccde5", "tree": "b9c119c7f8e935abb9466e9253b905a7f485d968", "parents": [ "45ee73a7fbe98cba2ccb007b60c027d27dfca1cb" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Mar 06 13:48:58 2015 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Mar 06 15:16:44 2015 -0800" }, "message": "Remove PTHREAD_ATTR_FLAG_MAIN_THREAD.\n\nMake this change because I think it is more reasonable to check stack info\nin pthread_getattr_np. I believe pthread_attr_t is not tied with any thread,\nand can\u0027t have a flag saying who using it is the main thread.\nThis change also helps refactor of g_thread_list_lock.\n\nBug: 19636317\nChange-Id: Iedbb85a391ac3e1849dd036d01445dac4bc63db9\n" }, { "commit": "5e2bd719d7dd19afe55f8d4f24366c0230e0e6c7", "tree": "14589a831046110c8cc6221cd1ee90226e899b4b", "parents": [ "04bbef377b9941bf6d67d9c515b6640a314fb032" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Feb 20 16:15:33 2015 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Mar 03 15:46:53 2015 -0800" }, "message": "Refactor pthread_key.cpp to be lock-free.\n\nChange-Id: I20dfb9d3cdc40eed10ea12ac34f03caaa94f7a49\n" }, { "commit": "19e246dda6772ffc532b1762cd7870d6c3b01c12", "tree": "93ab36fe44bc75b7366ad6d362ba25bc4ba4d994", "parents": [ "deab11acd7000573ad9c034e5bf9475aadbc5cbe" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Dec 18 14:22:09 2014 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Thu Jan 15 10:45:25 2015 -0800" }, "message": "Fix possible leak in pthread_detach.\n\nIf pthread_detach() is called while the thread is in pthread_exit(),\nit takes the risk that no one can free the pthread_internal_t.\nSo I add PTHREAD_ATTR_FLAG_ZOMBIE to detect this, maybe very rare, but\nboth glibc and netbsd libpthread have similar function.\n\nChange-Id: Iaa15f651903b8ca07aaa7bd4de46ff14a2f93835\n" }, { "commit": "ba8dfc2669d658dc340eb8f9c9b40ca074f05047", "tree": "d8e34d3dd322641c80c2605457378afd60d0bdfb", "parents": [ "597800f1d71590fc9ad51ce5227280586b2ceda5" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Jan 06 09:31:00 2015 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Jan 06 17:11:23 2015 -0800" }, "message": "Remove PTHREAD_ATTR_FLAG_USER_ALLOCATED_STACK.\n\nPatch for https://android-review.googlesource.com/#/c/120844/.\n\nChange-Id: Idca5ccd7b28e8f07f1d2d1b6e3bba6781b62f0e0\n" }, { "commit": "6a7aaf46759db32c6ed0eb953a4a230dc96af0d9", "tree": "89b9cbd40571d38d60c19730224c90efd0fb4ebf", "parents": [ "a2177c6ccb65104659543c895b753458abfadcab" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Mon Dec 22 19:17:33 2014 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Jan 02 14:33:48 2015 -0800" }, "message": "Reserve enough user request stack space in pthread_create.\n\nBug: 18830897\nChange-Id: I1ba4aaeaf66a7ff99c5d82ad45469011171b0a3b\n" }, { "commit": "8cf1b305670123aed7638d984ca39bfd22388440", "tree": "f8fc12a882822ca1ba41b68d84414e252faade9c", "parents": [ "c631bb215e29981222f19c092ded49c7c1f15845" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Wed Dec 03 21:36:24 2014 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Fri Dec 19 16:05:29 2014 -0800" }, "message": "Use mmap to create the pthread_internal_t\n\nAdd name to mmaped regions.\nAdd pthread benchmark code.\nAllocate pthread_internal_t on regular stack.\n\nBug: 16847284\nChange-Id: Id60835163bb0d68092241f1a118015b5a8f85069\n" }, { "commit": "8574a0670bf698b8d2f817098f9d825c4be03a68", "tree": "f01c1cd0cab5ff4bad74f9be90cd1dc0cc822513", "parents": [ "46d0b81a9465bf74fe44a881365c332dc6173f2f" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Mon Dec 01 22:37:56 2014 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Dec 02 10:39:25 2014 -0800" }, "message": "Use mmap to create the pthread_internal_t.\n\nBug: 16847284\nChange-Id: I488fa236f57aa3acb29b4ffbbab2fab51b0653be\n" }, { "commit": "5b8ceff5f87889e781c13305767e140afd28eb76", "tree": "ad95ccbfbc12b6a75ef56e6646cea4aee94b675c", "parents": [ "cc5f6543e3f91385b9a912438965b7e8265df54a" ], "author": { "name": "Nicolas Geoffray", "email": "ngeoffray@google.com", "time": "Wed Nov 26 11:53:44 2014 +0000" }, "committer": { "name": "Nicolas Geoffray", "email": "ngeoffray@google.com", "time": "Wed Nov 26 11:56:54 2014 +0000" }, "message": "Revert \"Use mmap to create the pthread_internal_t.\"\n\nUnfortunately, this change provokes random crashes for ART, and\nI have seen libc crashes on the device that might be related to it.\n\nReverting it fixes the ART crashes. there is unfortunately no\nstack trace for the crashes, but just a \"Segmentation fault\" message.\n\n\nThis reverts commit cc5f6543e3f91385b9a912438965b7e8265df54a.\n\nChange-Id: I68dca8e1e9b9edcce7eb84596e8db619e40e8052" }, { "commit": "cc5f6543e3f91385b9a912438965b7e8265df54a", "tree": "5176225b4d6b6f62ab590563022901042de44d8e", "parents": [ "eb3a5e026e65ea80b17580a71f4451f5cf5d6612" ], "author": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Nov 25 14:18:12 2014 -0800" }, "committer": { "name": "Yabin Cui", "email": "yabinc@google.com", "time": "Tue Nov 25 15:25:06 2014 -0800" }, "message": "Use mmap to create the pthread_internal_t.\n\nBug: 16847284\nChange-Id: Ic8c85f95afac1d8422ecb69674c688d1fecb6a44\n" }, { "commit": "04303f5a8ab9a992f3671d46b6ee2171582cbd61", "tree": "98d3997d33e93eeb91a2c2f331b57f35c180cbb2", "parents": [ "adc01348ee51a4ad678b1c277f85cbbed5c2e728" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Sep 18 16:11:59 2014 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Sep 19 17:37:06 2014 -0700" }, "message": "Add semaphore tests, fix sem_destroy.\n\nBug: https://code.google.com/p/android/issues/detail?id\u003d76088\nChange-Id: I4a0561b23e90312384d40a1c804ca64ee98f4066\n" }, { "commit": "a4831cb4a3f44b93788488db8ff9ea76613f0355", "tree": "eaa225c537d468307739bc48bbc86e241ada58e0", "parents": [ "ea66004c4a051e40c49314d808e7359634f6c641" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Sep 11 16:11:43 2014 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Sep 11 16:22:50 2014 -0700" }, "message": "Add pthread_gettid_np and re-expose __get_thread for LP32.\n\nA lot of third-party code calls the private __get_thread symbol,\noften as part of a backport of bionic\u0027s pthread_rwlock implementation.\nHopefully this will go away for LP64 (since you\u0027re guaranteed the\nreal implementation there), but there are still APIs that take a tid\nand no way to convert between a pthread_t and a tid. pthread_gettid_np\nis a public API for that. To aid the transition, make __get_thread\navailable again for LP32.\n\n(cherry-pick of 27efc48814b8153c55cbcd0af5d9add824816e69.)\n\nBug: 14079438\nChange-Id: I43fabc7f1918250d31d4665ffa4ca352d0dbeac1\n" }, { "commit": "40a521744825b6060960c296d5fb3da4c6593d94", "tree": "b9736886f5ca89daf4bf6cca05e090315a006dcd", "parents": [ "4ad5066e1de326e5db46df18eeade9a88bc11bec" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jul 30 14:48:10 2014 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jul 30 14:49:40 2014 -0700" }, "message": "Only wipe TLS for user-supplied stacks.\n\nBug: 16667988\nChange-Id: Id180ab2bc6713e1612386120a306db5bbf1d6046\n" }, { "commit": "7086ad6919feb2415c6027163f5c63323bcca27c", "tree": "f32c06d513e90ba1fcb8fe56f9cd055cccc7d01b", "parents": [ "d1bf37780d0bcaca3e6046171f958ebfea34bde1" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Jun 19 16:39:01 2014 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Jun 20 09:06:57 2014 -0700" }, "message": "Cache getpid.\n\nIn practice, with this implementation we never need to make a system call.\nWe get the main thread\u0027s tid (which is the same as our pid) back from\nthe set_tid_address system call we have to make during initialization.\nA new pthread will have the same pid as its parent, and a fork child\u0027s\nmain (and only) thread will have a pid equal to its tid, which we get for\nfree from the kernel before clone returns.\n\nThe only time we\u0027d actually have to make a getpid system call now is if\nwe take a signal during fork and the signal handler calls getpid. (That,\nor we call getpid in the dynamic linker while it\u0027s still dealing with its\nown relocations and hasn\u0027t even set up the main thread yet.)\n\nBug: 15387103\nChange-Id: I6d4718ed0a5c912fc75b5f738c49a023dbed5189\n" }, { "commit": "b30aff405a220495941f1673b0a5e66c4fa8b84c", "tree": "3bf667c009cf468c040ccaae6e81d468b4fd0254", "parents": [ "52f74322b1d72d57146107f32ee2c76c421bf4b1" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed May 28 19:35:33 2014 +0000" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed May 28 18:31:15 2014 -0700" }, "message": "Revert \"Revert \"Lose the hand-written futex assembler.\"\"\n\nThe problem with the original patch was that using syscall(3) means that\nerrno can be set, but pthread_create(3) was abusing the TLS errno slot as\na pthread_mutex_t for the thread startup handshake.\n\nThere was also a mistake in the check for syscall failures --- it should\nhave checked against -1 instead of 0 (not just because that\u0027s the default\nidiom, but also here because futex(2) can legitimately return values \u003e 0).\n\nThis patch stops abusing the TLS errno slot and adds a pthread_mutex_t to\npthread_internal_t instead. (Note that for LP64 sizeof(pthread_mutex_t) \u003e\nsizeof(uintptr_t), so we could potentially clobber other TLS slots too.)\n\nI\u0027ve also rewritten the LP32 compatibility stubs to directly reuse the\ncode from the .h file.\n\nThis reverts commit 75c55ff84ebfa686c7ae2cc8ee431c6a33bd46b4.\n\nBug: 15195455\nChange-Id: I6ffb13e5cf6a35d8f59f692d94192aae9ab4593d\n" }, { "commit": "1728b2396591853345507a063ed6075dfd251706", "tree": "7083cd234073afa5179b94b3d978550c890af90c", "parents": [ "bac795586bbc5dcbe886d8d781710f60c4c19d9b" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed May 14 10:02:03 2014 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed May 14 10:02:03 2014 -0700" }, "message": "Switch to g_ for globals.\n\nThat\u0027s what the Google style guide recommends, and we\u0027re starting\nto get a mix.\n\nChange-Id: Ib0c53a890bb5deed5c679e887541a715faea91fc\n" }, { "commit": "4b558f50a42c97d461f1dede5aaaae490ea99e2e", "tree": "c239a686131cdda28272c961ce98febbc0c772f8", "parents": [ "bef5016491eed41521f514d5c5528e2274689948" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Mar 04 15:58:02 2014 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Mar 07 16:49:46 2014 -0800" }, "message": "Rewrite the POSIX timer functions.\n\nThis is a much simpler implementation that lets the kernel\ndo as much as possible.\n\nCo-authored-by: Jörgen Strand \u003cjorgen.strand@sonymobile.com\u003e\nCo-authored-by: Snild Dolkow \u003csnild.dolkow@sonymobile.com\u003e\nChange-Id: Iad19f155de977667aea09410266d54e63e8a26bf\n" }, { "commit": "0e714a5b41451e84c5ded93a42c9a4b0a9440691", "tree": "86e6519cc723fb66a9432193da16b380255c22f7", "parents": [ "8e541b12237f597c186820b9e18b059b620bef54" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Mar 03 16:42:47 2014 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Mar 04 10:27:58 2014 -0800" }, "message": "Implement POSIX pthread_mutex_timedlock.\n\nThis replaces the non-standard pthread_mutex_lock_timeout_np, which we have\nto keep around on LP32 for binary compatibility.\n\nChange-Id: I098dc7cd38369f0c1bec1fac35687fbd27392e00\n" }, { "commit": "cef3faec0ea40fdfe58e425fd0be64f00de6a26d", "tree": "1f33cbb2de225d0d76d3eb53ef8857b9ecc79bbe", "parents": [ "1b2d214b62ad7bf46f37c79442a69d930aa29027" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Nov 19 16:52:24 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Nov 19 17:22:22 2013 -0800" }, "message": "Clean up pthread_internal_t.\n\nBug: 11755300\nChange-Id: Ib509e8c5ec6b23513aa78b5ac5141d7c34ce2dc8\n" }, { "commit": "36d6188f8cd8b948fb797f11d9620d63d0c2215a", "tree": "86b3162cac031151b2ae44b4d554fce4f9456147", "parents": [ "318e86ed887e04e593f3db9b84b402d5501ced9b" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Nov 19 13:31:58 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Nov 19 14:08:54 2013 -0800" }, "message": "Clean up forking and cloning.\n\nThe kernel now maintains the pthread_internal_t::tid field for us,\nand __clone was only used in one place so let\u0027s inline it so we don\u0027t\nhave to leave such a dangerous function lying around. Also rename\nfiles to match their content and remove some useless #includes.\n\nChange-Id: I24299fb4a940e394de75f864ee36fdabbd9438f9\n" }, { "commit": "877ec6d90418ff1d6597147d355a2229fdffae7e", "tree": "e475221a7fbff1564ad37548e920333c49cc5873", "parents": [ "f246c589d66e5dc0e3cddc3c37261fb0e3fc67e9" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Nov 15 17:40:18 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Nov 18 19:48:11 2013 -0800" }, "message": "Fix pthread_join.\n\nLet the kernel keep pthread_internal_t::tid updated, including\nacross forks and for the main thread. This then lets us fix\npthread_join to only return after the thread has really exited.\n\nAlso fix the thread attributes of the main thread so we don\u0027t\nunmap the main thread\u0027s stack (which is really owned by the\ndynamic linker and contains things like environment variables),\nwhich fixes crashes when joining with an exited main thread\nand also fixes problems reported publicly with accessing environment\nvariables after the main thread exits (for which I\u0027ve added a new\nunit test).\n\nIn passing I also fixed a bug where if the clone(2) inside\npthread_create(3) fails, we\u0027d unmap the child\u0027s stack and TLS (which\ncontains the mutex) and then try to unlock the mutex. Boom! It wasn\u0027t\nuntil after I\u0027d uploaded the fix for this that I came across a new\npublic bug reporting this exact failure.\n\nBug: 8206355\nBug: 11693195\nBug: https://code.google.com/p/android/issues/detail?id\u003d57421\nBug: https://code.google.com/p/android/issues/detail?id\u003d62392\nChange-Id: I2af9cf6e8ae510a67256ad93cad891794ed0580b\n" }, { "commit": "e48b68570d872ef7ece1d873c0ea298ea76393f3", "tree": "8c6d8721af2bfddd2c649a0d887a4a905358da04", "parents": [ "e39e3a77d6e8cf681718a572bc754fd64dfc87ff" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Nov 15 14:57:45 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Nov 15 14:57:45 2013 -0800" }, "message": "Clean up the pthread_create trampoline.\n\nBug: 8206355\nBug: 11693195\nChange-Id: I35cc024d5b6ebd19d1d2e45610db185addaf45df\n" }, { "commit": "70b24b1cc2a1a4436b1fea3f8b76616fdcb27224", "tree": "70c8e04ccd256eaeda3c7e516776f4c9bca25628", "parents": [ "05ec00bf62ac168c9787a3d0640879ab3e502fe5" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Nov 15 11:51:07 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Nov 15 14:41:19 2013 -0800" }, "message": "Switch pthread_create over to __bionic_clone.\n\nBug: 8206355\nBug: 11693195\nChange-Id: I04aadbc36c87e1b7e33324b9a930a1e441fbfed6\n" }, { "commit": "c3f114037dbf028896310609fd28cf2b3da99c4d", "tree": "e463aeb3604177cd93cea8a58936a8dce4b34ee5", "parents": [ "d0d0b52da375bfa85947a4257198791f615f18a8" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Oct 30 14:40:09 2013 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Oct 31 12:31:16 2013 -0700" }, "message": "\u003cpthread.h\u003e fixes and pthread cleanup.\n\n\u003cpthread.h\u003e was missing nonnull attributes, noreturn on pthread_exit,\nand had incorrect cv qualifiers for several standard functions.\n\nI\u0027ve also marked the non-standard stuff (where I count glibc rather\nthan POSIX as \"standard\") so we can revisit this cruft for LP64 and\ntry to ensure we\u0027re compatible with glibc.\n\nI\u0027ve also broken out the pthread_cond* functions into a new file.\n\nI\u0027ve made the remaining pthread files (plus ptrace) part of the bionic code\nand fixed all the warnings.\n\nI\u0027ve added a few more smoke tests for chunks of untested pthread functionality.\n\nWe no longer need the libc_static_common_src_files hack for any of the\npthread implementation because we long since stripped out the rest of\nthe armv5 support, and this hack was just to ensure that __get_tls in libc.a\nwent via the kernel if necessary.\n\nThis patch also finishes the job of breaking up the pthread.c monolith, and\nadds a handful of new tests.\n\nChange-Id: Idc0ae7f5d8aa65989598acd4c01a874fe21582c7\n" }, { "commit": "2b6e43e00ece68b3aba26d8f95f07cd9d9294ab4", "tree": "7a27d206fb0f2ef7b702c0c785827d607dd7216a", "parents": [ "8e6e7cdadbb4efdeefaa73fe94f7a9a1722745a7" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Oct 29 16:11:06 2013 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Oct 29 16:11:06 2013 -0700" }, "message": "Explain the sigprocmask in pthread_exit.\n\nAlso remove the SIGSEGV special case, which was probably because\nhand-written __exit_with_stack_teardown stubs used to try to cause\nSIGSEGV if the exit system call returned (which it never does, so\nthat dead code disappeared).\n\nAlso move the sigprocmask into the only case where it\u0027s necessary ---\nthe one where we unmap the stack that would be used by a signal\nhandler.\n\nChange-Id: Ie40d20c1ae2f5e7125131b6b492cba7a2c6d08e9\n" }, { "commit": "eb847bc8666842a3cfc9c06e8458ad1abebebaf0", "tree": "15e070fb706ba7599d5f402a8f934a7399541a6a", "parents": [ "0c17099a71cb5d1b36b7aab66fd7341b3fc9106d" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Oct 09 15:50:50 2013 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Oct 09 16:00:17 2013 -0700" }, "message": "Fix x86_64 build, clean up intermediate libraries.\n\nThe x86_64 build was failing because clone.S had a call to __thread_entry which\nwas being added to a different intermediate .a on the way to making libc.so,\nand the linker couldn\u0027t guarantee statically that such a relocation would be\npossible.\n\n ld: error: out/target/product/generic_x86_64/obj/STATIC_LIBRARIES/libc_common_intermediates/libc_common.a(clone.o): requires dynamic R_X86_64_PC32 reloc against \u0027__thread_entry\u0027 which may overflow at runtime; recompile with -fPIC\n\nThis patch addresses that by ensuring that the caller and callee end up in the\nsame intermediate .a. While I\u0027m here, I\u0027ve tried to clean up some of the mess\nthat led to this situation too. In particular, this removes libc/private/ from\nthe default include path (except for the DNS code), and splits out the DNS\ncode into its own library (since it\u0027s a weird special case of upstream NetBSD\ncode that\u0027s diverged so heavily it\u0027s unlikely ever to get back in sync).\n\nThere\u0027s more cleanup of the DNS situation possible, but this is definitely a\nstep in the right direction, and it\u0027s more than enough to get x86_64 building\ncleanly.\n\nChange-Id: I00425a7245b7a2573df16cc38798187d0729e7c4\n" }, { "commit": "50af69e8f326b2762a44d5fea2b118e7616e5d20", "tree": "46a173af8b370ce18d029c05c636bdc6fcc557ed", "parents": [ "322e7bce235ec6e3f82f65669423a1d9e997b51e" ], "author": { "name": "Brian Carlstrom", "email": "bdc@google.com", "time": "Fri Sep 13 16:34:43 2013 -0700" }, "committer": { "name": "Brian Carlstrom", "email": "bdc@google.com", "time": "Fri Sep 13 16:44:47 2013 -0700" }, "message": "Simplify main thread stack size initialization\n\nChange-Id: Iec09433d9de501031cce09dc75848a5e8f3d96bf\n" }, { "commit": "4cf13952175a793fed4646d7d7396364874b0909", "tree": "398598a421abf5790688df72da59cbce627d2e40", "parents": [ "e8bd8c2ed9c71c73682f29020456934c55f260e3" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Jul 19 16:42:27 2013 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Jul 19 16:42:27 2013 -0700" }, "message": "Move stuff only needed by pthread-timers.c into pthread-timers.c.\n\nChange-Id: I4915b3fff9c4f5a36b4f51027fb22019c11607b0\n" }, { "commit": "84114c8dd5b17efecf7988f263ce431208d7be5a", "tree": "81d4ae4e6a40c4255b2e76adeb2b68dab5932058", "parents": [ "40e7a87864fee75f7af2caf8e1972d0d764135cd" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jul 17 13:33:19 2013 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jul 17 13:33:19 2013 -0700" }, "message": "Improve stack overflow diagnostics (take 2).\n\nThis reverts commits eb1b07469f2b5a392dc1bfd8adc211aea8c72bc5 and\nd14dc3b87fbf80553f1cafa453816b7f11366627, and fixes the bug where\nwe were calling mmap (which might cause errno to be set) before\n__set_tls (which is required to implement errno).\n\nBug: 8557703\nChange-Id: I2c36d00240c56e156e1bb430d8c22a73a068b70c\n" }, { "commit": "eb1b07469f2b5a392dc1bfd8adc211aea8c72bc5", "tree": "b4cda0d4fa3935abd85b3ee178d35d0738e8a744", "parents": [ "5cf87951abd23b0b3ebf159e8aa06b02e39527ca" ], "author": { "name": "Guang Zhu", "email": "guangzhu@google.com", "time": "Wed Jul 17 03:16:04 2013 +0000" }, "committer": { "name": "Guang Zhu", "email": "guangzhu@google.com", "time": "Wed Jul 17 03:16:04 2013 +0000" }, "message": "Revert \"Clean up our alternate signal stacks.\"\n\nThis reverts commit 5cf87951abd23b0b3ebf159e8aa06b02e39527ca.\n\nChange-Id: Idd6ca7d80a018755da3bd315d91193723ce7f3bf\n" }, { "commit": "5cf87951abd23b0b3ebf159e8aa06b02e39527ca", "tree": "6eb2d74004073510406eb8363c81273533f71be4", "parents": [ "026867c7dcce7828212dcd4a61806146908e9039" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Jul 16 14:32:47 2013 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Jul 16 14:35:52 2013 -0700" }, "message": "Clean up our alternate signal stacks.\n\nBug: 8557703\nChange-Id: Ie93901dd1c29e9d3bf795b0f0400616d9ef08f75\n" }, { "commit": "0f020d18b138e24b1fe34074808e07ac412f35a4", "tree": "75e84d43c82a0e8013d7b1c34d4da1fec64e116c", "parents": [ "6a44d2271f372d0c65b05a5d3377bd00ce92824e" ], "author": { "name": "msg555", "email": "msg555@gmail.com", "time": "Thu Jun 06 14:59:28 2013 -0400" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jun 12 17:30:58 2013 -0700" }, "message": "Handles spurious wake-ups in pthread_join()\n\nRemoved \u0027join_count\u0027 from pthread_internal_t and switched to using the flag\nPTHREAD_ATTR_FLAG_JOINED to indicate if a thread is being joined. Combined with\na switch to a while loop in pthread_join, this fixes spurious wake-ups but\nprevents a thread from being joined multiple times. This is fine for\ntwo reasons:\n\n1) The pthread_join specification allows for undefined behavior when multiple\n threads try to join a single thread.\n\n2) There is no thread safe way to allow multiple threads to join a single\n thread with the pthread interface. The second thread calling pthread_join\n could be pre-empted until the thread is destroyed and its handle reused for\n a different thread. Therefore multi-join is always an error.\n\nBug: https://code.google.com/p/android/issues/detail?id\u003d52255\nChange-Id: I8b6784d47620ffdcdbfb14524e7402e21d46c5f7\n" }, { "commit": "36fa67bcdd90f18a3c68f8637ae836762407fa51", "tree": "67c47601c9dd09bf8fc97d15e872c264752477da", "parents": [ "c5bfb62433751579e86d5fd24cddf3ae28a49b43" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jun 05 17:51:20 2013 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Jun 05 17:58:08 2013 -0700" }, "message": "Ensure header files using __BEGIN_DECLS include sys/cdefs.h.\n\nWe keep fixing these one-by-one; let\u0027s fix them all at once.\n\nFound thus:\n\n find . -name *.h | xargs grep -L sys/cdefs.h | xargs grep -l BEGIN_DECL | xargs grep -L sys/types\n\nChange-Id: I188842aa2484dc6176e96556d57c38a0f785b59b\n" }, { "commit": "9d23e04c43dbb8480bea8be28b8a2f37423bec49", "tree": "e403288cf01ce40d6f330da09a317c7ad6dc2cb2", "parents": [ "081318e3550b1a8d8384d98d7c94527681691120" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Feb 15 19:21:51 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 19 12:21:41 2013 -0800" }, "message": "Fix pthreads functions that should return ESRCH.\n\nimgtec pointed out that pthread_kill(3) was broken, but most of the\nother functions that ought to return ESRCH for invalid/exited threads\nwere equally broken.\n\nChange-Id: I96347f6195549aee0c72dc39063e6c5d06d2e01f\n" }, { "commit": "40eabe24e4e3ae8ebe437f1f4e43cf39cbba2e9e", "tree": "19eef7fcbba616a52bd37f4c83ece3e918b25445", "parents": [ "3e3b239d2be0c5a92e91385f4f10b8f7da8d2837" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Feb 14 18:59:37 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Feb 15 12:08:59 2013 -0800" }, "message": "Fix the pthread_setname_np test.\n\nFix the pthread_setname_np test to take into account that emulator kernels are\nso old that they don\u0027t support setting the name of other threads.\n\nThe CLONE_DETACHED thread is obsolete since 2.5 kernels.\n\nRename kernel_id to tid.\n\nFix the signature of __pthread_clone.\n\nClean up the clone and pthread_setname_np implementations slightly.\n\nChange-Id: I16c2ff8845b67530544bbda9aa6618058603066d\n" }, { "commit": "6d339182070b035ca94f19bc37c94f4d9813c374", "tree": "3745d21667b8c856276c85cf0864d6ab483f83cf", "parents": [ "4912782c6af7169686acc9553fb0bb33251b0d0d" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 12 16:36:04 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 12 16:36:04 2013 -0800" }, "message": "Simplify pthread_create, using more public API.\n\nChange-Id: I08e65ba88ed01436223e4e528631c9e41ec0e7f4\n" }, { "commit": "3e898476c7230b60a0f76968e64ff25f475b48c0", "tree": "0f876aeb565b7e7ac627a6305a355c40173912f2", "parents": [ "fcaf4e9f9b735e053469c7ecbf63584e10fd67a7" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 12 16:40:24 2013 +0000" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 12 15:27:18 2013 -0800" }, "message": "Revert \"Revert \"More pthreads cleanup.\"\"\n\nThis reverts commit 6f94de3ca49e4ea147b1c59e5818fa175846518f\n\n(Doesn\u0027t try to increase the number of TLS slots; that leads to\nan inability to boot. Adds more tests.)\n\nChange-Id: Ia7d25ba3995219ed6e686463dbba80c95cc831ca\n" }, { "commit": "6f94de3ca49e4ea147b1c59e5818fa175846518f", "tree": "9a2c45004114f53c4cf1f9468a58fe54c58320fa", "parents": [ "2a1bb4e64677b9abbc17173c79768ed494565047" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Feb 12 06:06:22 2013 +0000" }, "committer": { "name": "Gerrit Code Review", "email": "noreply-gerritcodereview@google.com", "time": "Tue Feb 12 06:06:22 2013 +0000" }, "message": "Revert \"More pthreads cleanup.\"\n\nThis reverts commit 2a1bb4e64677b9abbc17173c79768ed494565047\n\nChange-Id: Ia443d0748015c8e9fc3121e40e68258616767b51\n" }, { "commit": "2a1bb4e64677b9abbc17173c79768ed494565047", "tree": "3e843fd4277f2bdc502511bd5e4ee539887c14f1", "parents": [ "2d3e72336e76180fb00822386da4f14203d117ce" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Feb 11 12:34:03 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Feb 11 14:56:39 2013 -0800" }, "message": "More pthreads cleanup.\n\nPOSIX says pthread_create returns EAGAIN, not ENOMEM.\n\nAlso pull pthread_attr_t functions into their own file.\n\nAlso pull pthread_setname_np into its own file.\n\nAlso remove unnecessary #includes from pthread_key.cpp.\n\nAlso account for those pthread keys used internally by bionic,\nso they don\u0027t count against the number of keys available to user\ncode. (They do with glibc, but glibc\u0027s limit is the much more\ngenerous 1024.)\n\nAlso factor out the common errno-restoring idiom to reduce gotos.\n\nBug: 6702535\nChange-Id: I555e66efffcf2c1b5a2873569e91489156efca42\n" }, { "commit": "44b53ad6818de344e0b499ad8fdbb21fcb0ff2b6", "tree": "9a2c45004114f53c4cf1f9468a58fe54c58320fa", "parents": [ "8397cdba9424febeaed4068829a5b0174ee1138c" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Feb 11 20:18:47 2013 +0000" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Feb 11 12:20:33 2013 -0800" }, "message": "Revert \"Revert \"Pull the pthread_key_t functions out of pthread.c.\"\"\n\nThis reverts commit 6260553d48f6fd87ca220270bea8bafdde5726ec\n\n(Removing the accidental libm/Android.mk change.)\n\nChange-Id: I6cddd9857c31facc05636e8221505b3d2344cb75\n" }, { "commit": "6260553d48f6fd87ca220270bea8bafdde5726ec", "tree": "9c2cb4dfb830d7918a0d0b8ecfbf25bf87bcc453", "parents": [ "ad59322ae432d11ff36dcf046016af8cfe45fbe4" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Feb 11 20:18:16 2013 +0000" }, "committer": { "name": "Gerrit Code Review", "email": "noreply-gerritcodereview@google.com", "time": "Mon Feb 11 20:18:16 2013 +0000" }, "message": "Revert \"Pull the pthread_key_t functions out of pthread.c.\"\n\nThis reverts commit ad59322ae432d11ff36dcf046016af8cfe45fbe4\n\nsomehow my unfinished libm/Android.mk change got into here.\n\nChange-Id: I46be626c5269d60fb1ced9862f2ebaa380b4e0af\n" }, { "commit": "ad59322ae432d11ff36dcf046016af8cfe45fbe4", "tree": "04ea728ddd656079ca4b9f228b32c371370be947", "parents": [ "9a9bb243b50be5e3910b8edad72327bc216e72d0" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Fri Feb 08 15:46:37 2013 -0800" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Feb 11 12:00:48 2013 -0800" }, "message": "Pull the pthread_key_t functions out of pthread.c.\n\nThis was originally motivated by noticing that we were setting the\nwrong bits for the well-known tls entries. That was a harmless bug\nbecause none of the well-known tls entries has a destructor, but\nit\u0027s best not to leave land mines lying around.\n\nAlso add some missing POSIX constants, a new test, and fix\npthread_key_create\u0027s return value when we hit the limit.\n\nChange-Id: Ife26ea2f4b40865308e8410ec803b20bcc3e0ed1\n" }, { "commit": "4f251bee5d51228217c1bf4dfc9219f3058bd3ed", "tree": "3d0f1faef6a38e494f8157e467e7fc0f876f6cd4", "parents": [ "064f862d557ab741575dfae479499a07ca0ab742" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Nov 01 16:33:29 2012 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Thu Nov 01 17:27:07 2012 -0700" }, "message": "Don\u0027t corrupt the thread list if the main thread exits.\n\n...and don\u0027t pass a non-heap pointer to free(3), either.\n\nThis patch replaces the \"node** prev\" with the clearer \"node* prev\"\nstyle and fixes the null pointer dereference in the old code. That\u0027s\nnot sufficient to fix the reporter\u0027s bug, though. The pthread_internal_t*\nfor the main thread isn\u0027t heap-allocated --- __libc_init_tls causes a\npointer to a statically-allocated pthread_internal_t to be added to\nthe thread list.\n\nBug: http://code.google.com/p/android/issues/detail?id\u003d37410\nChange-Id: I112b7f22782fc789d58f9c783f7b323bda8fb8b7\n" }, { "commit": "ae8eb74675722b57ab66a51f1d6f4f250137bb23", "tree": "7a5b7bf7eed38b5a7a5129cd1f261e66277381b0", "parents": [ "33c4a3adde3cf1753af0661f48ce86358e105277" ], "author": { "name": "Xi Wang", "email": "xi.wang@gmail.com", "time": "Sat Oct 27 02:02:01 2012 -0400" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Mon Oct 29 08:22:13 2012 -0700" }, "message": "Fix a potential NULL pointer dereference in _init_thread().\n\nThe first NULL pointer check against `attr\u0027 suggests that `attr\u0027 can\nbe NULL. Then later `attr\u0027 is directly dereferenced, suggesting the\nopposite.\n\n if (attr \u003d\u003d NULL) {\n ...\n } else {\n ...\n }\n ...\n if (attr-\u003estack_base \u003d\u003d ...) { ... }\n\nThe public API pthread_create(3) allows NULL, and interprets it as \"default\".\nOur implementation actually swaps in a pointer to the global default\npthread_attr_t, so we don\u0027t need any NULL checks in _init_thread. (The other\ninternal caller passes its own pthread_attr_t.)\n\nChange-Id: I0a4e79b83f5989249556a07eed1f2887e96c915e\nSigned-off-by: Xi Wang \u003cxi.wang@gmail.com\u003e\n" }, { "commit": "5419b9474753d25dff947c7740532f86d130c0be", "tree": "4d746cfc20a1d3b5886f691ed1a49ddf34e2df78", "parents": [ "a9944cfe9e152ca46afb0a77300ec5a2a1a24e64" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Oct 16 15:54:46 2012 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue Oct 16 17:58:17 2012 -0700" }, "message": "Make dlerror(3) thread-safe.\n\nI gave up trying to use the usual thread-local buffer idiom; calls to\ncalloc(3) and free(3) from any of the \"dl\" functions -- which live in\nthe dynamic linker -- end up resolving to the dynamic linker\u0027s stubs.\nI tried to work around that, but was just making things more complicated.\nThis alternative costs us a well-known TLS slot (instead of the\ndynamically-allocated TLS slot we\u0027d have used otherwise, so no difference\nthere), plus an extra buffer inside every pthread_internal_t.\n\nBug: 5404023\nChange-Id: Ie9614edd05b6d1eeaf7bf9172792d616c6361767\n" }, { "commit": "bfeab1bbe7e8d0c08b7e3f46aedab64e3b2bf706", "tree": "d7ee2fec0079394cfacebdc632036a53f079c011", "parents": [ "ff4608ae0cb78caa05b83e109fc2633682958089" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Sep 05 17:47:37 2012 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Wed Sep 05 17:47:37 2012 -0700" }, "message": "Don\u0027t corrupt the thread list in static executables.\n\nSeveral previous changes conspired to make a mess of the thread list\nin static binaries. This was most obvious when trying to call\npthread_key_delete(3) on the main thread.\n\nBug: http://code.google.com/p/android/issues/detail?id\u003d36893\nChange-Id: I2a2f553114d8fb40533c481252b410c10656da2e\n" }, { "commit": "9c3eca7bcee694e6a477a7d50065f11cf1e805bb", "tree": "abb36a7c1f2f13d54caad0f5a8175e47dd153730", "parents": [ "ff0d1ce4dfceb35a65440259c3f325106fc9e39b", "6cf3c7c50a4b9bc8425562c542f6a3d320db70c3" ], "author": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue May 08 13:26:28 2012 -0700" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue May 08 13:26:28 2012 -0700" }, "message": "resolved conflicts for merge of 6cf3c7c5 to jb-dev-plus-aosp\n\nChange-Id: Ib22a8ae1ef63bf9b9c82ce6e22afd188ba7c2806\n" }, { "commit": "d0c884d3595ecca03c3e70de9909c090cd5f9cae", "tree": "a15fbefbe3ffcf1fc111ec284a1c62dcf3beddc6", "parents": [ "e3bc7192ec4254bed00eb8b352735665c6072995" ], "author": { "name": "Pierre Peiffer", "email": "pierre.peiffer@stericsson.com", "time": "Wed Feb 22 16:40:15 2012 +0100" }, "committer": { "name": "Elliott Hughes", "email": "enh@google.com", "time": "Tue May 08 10:54:51 2012 -0700" }, "message": "Let pthread_create fail if schedparam can\u0027t be set\n\nThe creation of a thread succeeds even if the requested scheduling\nparameters can not be set. This is not POSIX compliant, and even\nworse, it leads to a wrong behavior. Let pthread_create() fail in this\ncase.\n\nChange-Id: Ice66e2a720975c6bde9fe86c2cf8f649533a169c\nSigned-off-by: Christian Bejram \u003cchristian.bejram@stericsson.com\u003e\n" }, { "commit": "1a78fbb5c8228e4aea2a516818828b76044310f2", "tree": "4de785711dbfa8923d721f9180473328b6228a5c", "parents": [ "d5099016f70b5acbfeb969787687099df703a4f4" ], "author": { "name": "Evgeniy Stepanov", "email": "eugenis@google.com", "time": "Thu Mar 22 18:01:53 2012 +0400" }, "committer": { "name": "Evgeniy Stepanov", "email": "eugenis@google.com", "time": "Fri Apr 13 00:08:11 2012 +0400" }, "message": "Initialize TLS before any application code is run.\n\nSince e19d702b8e33, dlsym and friends use recursive mutexes that\nrequire the current thread id, which is not available before the libc\nconstructor. This prevents us from using dlsym() in .preinit_array.\n\nThis change moves TLS initialization from libc constructor to the earliest\npossible point - immediately after linker itself is relocated. As a result,\npthread_internal_t for the initial thread is available from the start.\n\nAs a bonus, values stored in TLS in .preinit_array are not lost when libc is\ninitialized.\n\nChange-Id: Iee5a710ee000173bff63e924adeb4a4c600c1e2d\n" }, { "commit": "4f086aeb4aa06e13079b7fec71a8178ceeacf318", "tree": "a6b0fbce5c6d0f507a244ef4f87ec648b7c35a76", "parents": [ "e73a571fd9b60315681eacbef5857f0f0b1ece18" ], "author": { "name": "Matt Fischer", "email": "matt.fischer@garmin.com", "time": "Fri Jun 25 14:36:39 2010 -0500" }, "committer": { "name": "Jean-Baptiste Queru", "email": "jbq@google.com", "time": "Fri Jul 23 13:10:16 2010 -0700" }, "message": "Implemented pthread_atfork()\n\nChange-Id: Ie6c0bf593315d3507b3c4a6c8903a74a1fa053db\n" }, { "commit": "3a654b1e04d4275ae315cfe1b196998acf10052c", "tree": "79ce08ac573c6c5bd7d7a65f7fdc24237c00452a", "parents": [ "03eabfe65e1e2c36f4d26c78a730fa19a3bdada3" ], "author": { "name": "David \u0027Digit\u0027 Turner", "email": "digit@google.com", "time": "Wed Jun 03 19:32:37 2009 +0200" }, "committer": { "name": "David \u0027Digit\u0027 Turner", "email": "digit@google.com", "time": "Wed Jun 03 19:32:37 2009 +0200" }, "message": "Revert \"Fix the C library initialization to avoid calling static C++ constructors twice.\"\n\nThis reverts commit 03eabfe65e1e2c36f4d26c78a730fa19a3bdada3.\n" }, { "commit": "03eabfe65e1e2c36f4d26c78a730fa19a3bdada3", "tree": "b965ea27e54b0833639227c619f6e35647c92510", "parents": [ "0353195f344666256dba474a15c9ba22cf0cccc9" ], "author": { "name": "David \u0027Digit\u0027 Turner", "email": "digit@google.com", "time": "Thu May 28 15:54:03 2009 +0200" }, "committer": { "name": "David \u0027Digit\u0027 Turner", "email": "digit@google.com", "time": "Tue Jun 02 23:27:44 2009 +0200" }, "message": "Fix the C library initialization to avoid calling static C++ constructors twice.\n\nThe problem was due to the fact that, in the case of dynamic executables,\nthe dynamic linker calls the DT_PREINIT_ARRAY, DT_INIT and DT_INIT_ARRAY\nconstructors when loading shared libraries and dynamic executables,\n*before* calling the executable\u0027s entry point (i.e. arch-$ARCH/bionic/crtbegin_dynamic.c)\nwhich in turns call __libc_init() in libc.so, as defined by bionic/libc_init_dynamic.c\n\nThe latter did call these constructors array again, mistakenly.\n\nThe patch also updates the documentation of many related functions.\n\nAlso adds a new section to linker/README.TXT explaining restrictions on\nC library usage.\n\nThe patch has been tested on a Dream for stability issues with\nproprietary blobs:\n\n- H264 decoding works\n- Camera + Video recording works\n- GPS works\n- Sensors work\n\nThe tests in system/extra/tests/bionic/libc/common/test_static_cpp_mutex.cpp has been\nrun and shows the static C++ constructor being called only once.\n" }, { "commit": "1dc9e472e19acfe6dc7f41e429236e7eef7ceda1", "tree": "3be0c520fae17689bbf5584e1136fb820caef26f", "parents": [ "1767f908af327fa388b1c66883760ad851267013" ], "author": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Tue Mar 03 19:28:35 2009 -0800" }, "committer": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Tue Mar 03 19:28:35 2009 -0800" }, "message": "auto import from //depot/cupcake/@135843\n" }, { "commit": "1767f908af327fa388b1c66883760ad851267013", "tree": "4b825dc642cb6eb9a060e54bf8d69288fbee4904", "parents": [ "a799b53f10e5a6fd51fef4436cfb7ec99836a516" ], "author": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Tue Mar 03 18:28:13 2009 -0800" }, "committer": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Tue Mar 03 18:28:13 2009 -0800" }, "message": "auto import from //depot/cupcake/@135843\n" }, { "commit": "4e468ed2eb86a2406e14f1eca82072ee501d05fd", "tree": "4e05b3c66eef86531e464521a3bf96a1864d4bf5", "parents": [ "a27d2baa0c1a2ec70f47ea9199b1dd6762c8a349" ], "author": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Wed Dec 17 18:03:48 2008 -0800" }, "committer": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Wed Dec 17 18:03:48 2008 -0800" }, "message": "Code drop from //branches/cupcake/...@124589\n" }, { "commit": "a27d2baa0c1a2ec70f47ea9199b1dd6762c8a349", "tree": "defd1cc07d16ad2f3b21154114e092d11c94c5bb", "parents": [], "author": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Tue Oct 21 07:00:00 2008 -0700" }, "committer": { "name": "The Android Open Source Project", "email": "initial-contribution@android.com", "time": "Tue Oct 21 07:00:00 2008 -0700" }, "message": "Initial Contribution\n" } ] }