[openib-general] the valgrind support of libibverbs/libmthca is not full

Dotan Barak dotanb at dev.mellanox.co.il
Sat Jan 6 23:26:38 PST 2007


Hi Roland and welcome back ....

Roland Dreier wrote:
>  > For the following verbs (and I'm sure that this is not the full list) 
>  > there is a memory warning from valgrind:
>  > ibv_query_device
>  > ibv_dealloc_pd
>  > ibv_create_cq
>
> Can you send the full warnings you get?
>
> Thanks
>   
Here are some of the messages that i got when i executed gen2_basic 
under valgrind:

==10826== Use of uninitialised value of size 8
==10826==    at 0x39A843E79A: _itoa_word (in /lib64/tls/libc-2.3.4.so)
==10826==    by 0x39A844196A: vfprintf (in /lib64/tls/libc-2.3.4.so)
==10826==    by 0x39A8462433: vsnprintf (in /lib64/tls/libc-2.3.4.so)
==10826==    by 0x39A84481E0: snprintf (in /lib64/tls/libc-2.3.4.so)
==10826==    by 0x4C3DA3C: mthca_query_device (verbs.c:63)
==10826==    by 0x404363: hca_4 (test_hca.c:192)
==10826==    by 0x40571C: test_hca (test_hca.c:368)
==10826==    by 0x403111: main (main.c:135)

==10853== 72 bytes in 1 blocks are definitely lost in loss record 8 of 13
==10853==    at 0x4904B7E: malloc (vg_replace_malloc.c:149)
==10853==    by 0x4C3DA82: mthca_alloc_pd (verbs.c:83)
==10853==    by 0x4C3BC15: mthca_alloc_context (mthca.c:179)
==10853==    by 0x4A32536: ibv_open_device (device.c:126)
==10853==    by 0x401FF4: open_hca (main.c:214)
==10853==    by 0x40E4E7: cq_7 (test_cq.c:529)
==10853==    by 0x4108A0: test_cq (test_cq.c:790)
==10853==    by 0x403196: main (main.c:147)


==10855== Syscall param write(buf) points to uninitialised byte(s)
==10855==    at 0x39A84B9302: __write_nocancel (in /lib64/tls/libc-2.3.4.so)
==10855==    by 0x4A30B2E: ibv_cmd_reg_mr (cmd.c:245)
==10855==    by 0x4C3DBA3: __mthca_reg_mr (verbs.c:134)
==10855==    by 0x4C3DCE9: mthca_create_cq (mthca.h:259)
==10855==    by 0x4A34174: ibv_create_cq (verbs.c:247)
==10855==    by 0x40E513: cq_7 (test_cq.c:533)
==10855==    by 0x4108A0: test_cq (test_cq.c:790)
==10855==    by 0x403196: main (main.c:147)
==10855==  Address 0x7FF0006A8 is on thread 1's stack


==10855== Syscall param write(buf) points to uninitialised byte(s)
==10855==    at 0x39A84B9302: __write_nocancel (in /lib64/tls/libc-2.3.4.so)
==10855==    by 0x4A30D43: ibv_cmd_create_cq (cmd.c:315)
==10855==    by 0x4C3DDED: mthca_create_cq (verbs.c:233)
==10855==    by 0x4A34174: ibv_create_cq (verbs.c:247)
==10855==    by 0x40E513: cq_7 (test_cq.c:533)
==10855==    by 0x4108A0: test_cq (test_cq.c:790)
==10855==    by 0x403196: main (main.c:147)
==10855==  Address 0x7FF000748 is on thread 1's stack


If you need some more info, just tell me ...

thanks
Dotan






More information about the general mailing list