Changes between Version 1 and Version 2 of helgrind


Ignore:
Timestamp:
Apr 2, 2015 6:02:56 AM (9 years ago)
Author:
ming
Comment:

--

Legend:

Unmodified
Added
Removed
Modified

  • helgrind

    v1 v2  
    11This wiki describes the issue from ticket #1820 (Helgrind thread error detector test and analysis) 
    22 
    3 Helgrind issue !#1 
    4  
    5 Description: Data race over possible concurrent read/write access for thread's quit flag.[[br]] 
    6 One thread tries to write the quit flag while another tries to read it.[[br]] 
    7 Status: false positive, this is intentional behaviour. 
    8  
    9 {{{ 
    10 ==26641== ---Thread-Announcement------------------------------------------ 
    11 ==26641== 
    12 ==26641== Thread #6 was created 
    13 ==26641==    at 0x5953FCE: clone (clone.S:74) 
    14 ==26641==    by 0x5352199: do_clone.constprop.3 (createthread.c:75) 
    15 ==26641==    by 0x53538BA: pthread_create@@GLIBC_2.2.5 (createthread.c:245) 
    16 ==26641==    by 0x4C30C90: pthread_create_WRK (hg_intercepts.c:269) 
    17 ==26641==    by 0x56704B: pj_thread_create (os_core_unix.c:616) 
    18 ==26641==    by 0x49CC57: pjmedia_endpt_create (endpoint.c:169) 
    19 ==26641==    by 0x43102A: pjsua_media_subsys_init (pjsua_media.c:80) 
    20 ==26641==    by 0x42B576: pjsua_init (pjsua_core.c:1058) 
    21 ==26641==    by 0x40800A: app_init (pjsua_app.c:1346) 
    22 ==26641==    by 0x408F1A: pjsua_app_init (pjsua_app.c:1881) 
    23 ==26641==    by 0x405AA9: main_func (main.c:108) 
    24 ==26641==    by 0x568914: pj_run_app (os_core_unix.c:1930) 
    25 ==26641== 
    26 ==26641== Possible data race during write of size 4 at 0x8182E8 by thread #1 
    27 ==26641== Locks held: none 
    28 ==26641==    at 0x42AA5E: pjsua_stop_worker_threads (pjsua_core.c:730) 
    29 ==26641==    by 0x42C753: pjsua_destroy2 (pjsua_core.c:1548) 
    30 ==26641==    by 0x42CDC9: pjsua_destroy (pjsua_core.c:1775) 
    31 ==26641==    by 0x409291: app_destroy (pjsua_app.c:2011) 
    32 ==26641==    by 0x409304: pjsua_app_destroy (pjsua_app.c:2035) 
    33 ==26641==    by 0x405ADF: main_func (main.c:116) 
    34 ==26641==    by 0x568914: pj_run_app (os_core_unix.c:1930) 
    35 ==26641==    by 0x405B32: main (main.c:129) 
    36 ==26641== 
    37 ==26641== This conflicts with a previous read of size 4 by thread #7 
    38 ==26641== Locks held: none 
    39 ==26641==    at 0x42A978: worker_thread (pjsua_core.c:691) 
    40 ==26641==    by 0x566E30: thread_main (os_core_unix.c:523) 
    41 ==26641==    by 0x4C30E26: mythread_wrapper (hg_intercepts.c:233) 
    42 ==26641==    by 0x5353181: start_thread (pthread_create.c:312) 
    43 ==26641==    by 0x595400C: clone (clone.S:111) 
    44 }}} 
    45  
    46 Helgrind issue !#2 
    47  
    48 Lock order inversion involving rwmutex_lock_read()[[br]] 
     31. Data race over possible concurrent read/write access for thread's quit flag.[[br]] 
     4  Description: One thread tries to write the quit flag while another tries to read it.[[br]] 
     5  Status: false positive, this is intentional behaviour. 
     6 
     7  {{{ 
     8  ==26641== ---Thread-Announcement------------------------------------------ 
     9  ==26641== 
     10  ==26641== Thread #6 was created 
     11  ==26641==    at 0x5953FCE: clone (clone.S:74) 
     12  ==26641==    by 0x5352199: do_clone.constprop.3 (createthread.c:75) 
     13  ==26641==    by 0x53538BA: pthread_create@@GLIBC_2.2.5 (createthread.c:245) 
     14  ==26641==    by 0x4C30C90: pthread_create_WRK (hg_intercepts.c:269) 
     15  ==26641==    by 0x56704B: pj_thread_create (os_core_unix.c:616) 
     16  ==26641==    by 0x49CC57: pjmedia_endpt_create (endpoint.c:169) 
     17  ==26641==    by 0x43102A: pjsua_media_subsys_init (pjsua_media.c:80) 
     18  ==26641==    by 0x42B576: pjsua_init (pjsua_core.c:1058) 
     19  ==26641==    by 0x40800A: app_init (pjsua_app.c:1346) 
     20  ==26641==    by 0x408F1A: pjsua_app_init (pjsua_app.c:1881) 
     21  ==26641==    by 0x405AA9: main_func (main.c:108) 
     22  ==26641==    by 0x568914: pj_run_app (os_core_unix.c:1930) 
     23  ==26641== 
     24  ==26641== Possible data race during write of size 4 at 0x8182E8 by thread #1 
     25  ==26641== Locks held: none 
     26  ==26641==    at 0x42AA5E: pjsua_stop_worker_threads (pjsua_core.c:730) 
     27  ==26641==    by 0x42C753: pjsua_destroy2 (pjsua_core.c:1548) 
     28  ==26641==    by 0x42CDC9: pjsua_destroy (pjsua_core.c:1775) 
     29  ==26641==    by 0x409291: app_destroy (pjsua_app.c:2011) 
     30  ==26641==    by 0x409304: pjsua_app_destroy (pjsua_app.c:2035) 
     31  ==26641==    by 0x405ADF: main_func (main.c:116) 
     32  ==26641==    by 0x568914: pj_run_app (os_core_unix.c:1930) 
     33  ==26641==    by 0x405B32: main (main.c:129) 
     34  ==26641== 
     35  ==26641== This conflicts with a previous read of size 4 by thread #7 
     36  ==26641== Locks held: none 
     37  ==26641==    at 0x42A978: worker_thread (pjsua_core.c:691) 
     38  ==26641==    by 0x566E30: thread_main (os_core_unix.c:523) 
     39  ==26641==    by 0x4C30E26: mythread_wrapper (hg_intercepts.c:233) 
     40  ==26641==    by 0x5353181: start_thread (pthread_create.c:312) 
     41  ==26641==    by 0x595400C: clone (clone.S:111) 
     42  }}} 
     43 
     442. Lock order inversion involving rwmutex_lock_read() 
    4945Both threads try to acquire read locks, which should be ok. However, on some platforms, read write mutex may be implemented as some regular mutex, and deadlock can occur. 
    5046 
    51 {{{ 
     47  {{{ 
    5248==26641== Thread #7: lock order "0x5EDC798 before 0x5E25638" violated 
    5349(lock read should be ok) 
    … …  
    110106}}} 
    111107 
    112 Helgrind issue !#3 
    113  
    114 Pulse Audio related[[br]] 
    115 Status: outside PJSIP's scope[[br]] 
    116 TODO: create a suppression file for this 
     1083. Pulse Audio related[[br]] 
     109  Status: outside PJSIP's scope[[br]] 
     110  TODO: create a suppression file for this 
    117111 
    118112{{{ 
    … …  
    149143}}} 
    150144 
    151 Helgrind issue !#4 
    152  
    153 Data race over access to g_last_thread variable.[[br]] 
     1454. Data race over access to g_last_thread variable. 
    154146Status: won't fix. The variable is only used for information purpose only, while mutex locking for logging will be quite expensive. 
    155147 
    … …  
    208200}}} 
    209201 
    210 Helgrind issue !#5 
    211  
    212 Data race over cp->used_size variable in pool_caching.[[br]] 
     2025. Data race over cp->used_size variable in pool_caching. 
    213203Status: won't fix, the variable is for statistics purpose only. 
    214204 
    … …  
    251241}}} 
    252242 
    253 Helgrind issue !#6 
    254  
    255 Data race over tcp->pending_connect variable[[br]] 
     2436. Data race over tcp->pending_connect variable[[br]] 
    256244Status: should be safe. A modification to suppress this would be desirable. 
    257245 
    … …  
    340328==5950==    by 0x408DCF: app_init (pjsua_app.c:1813) 
    341329}}} 
     330 
     3317. Destroying a locked mutex in sip dialog 
     332Status: won't fix. When calling pj_mutex_destroy() there is actually no requirement that the mutex should be unlocked. If the mutex destroy fails, pj_mutex_destroy() will automatically retry several times by unlocking the mutex first. 
     333{{{ 
     334==27298== Thread #9 unlocked an invalid lock at 0x6B1B408 
     335==27298==    at 0x4C325C0: pthread_mutex_unlock (hg_intercepts.c:632) 
     336==27298==    by 0x578334: pj_mutex_destroy (os_core_unix.c:1413) 
     337==27298==    by 0x4F8D7A: destroy_dialog (sip_dialog.c:114) 
     338==27298==    by 0x4FA4C9: unregister_and_destroy_dialog (sip_dialog.c:760) 
     339==27298==    by 0x4FA94B: pjsip_dlg_dec_lock (sip_dialog.c:896) 
     340==27298==    by 0x4FCD4A: pjsip_dlg_on_tsx_state (sip_dialog.c:2030) 
     341==27298==    by 0x4FD674: mod_ua_on_tsx_state (sip_ua_layer.c:178) 
     342==27298==    by 0x4F4771: tsx_set_state (sip_transaction.c:1210) 
     343==27298==    by 0x4F8103: tsx_on_state_completed_uas (sip_transaction.c:3150) 
     344==27298==    by 0x4F4591: tsx_timer_callback (sip_transaction.c:1153) 
     345==27298==    by 0x58808F: pj_timer_heap_poll (timer.c:643) 
     346==27298==    by 0x4DF50A: pjsip_endpt_handle_events2 (sip_endpoint.c:712) 
     347}}}