extend support for CUDA features #3
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
add error checking for cuMemFree
patch from Huy Nguyen <[email protected]>
…troduce --cuda_mem_hints
…NT_MANAGED_ACCESS as proxy to ATS
…. use separate error string for the different pp_init_ functions.
drossetti
pushed a commit
that referenced
this pull request
May 12, 2021
The sizeof 'struct ibv_qp' allocated by ibv_create_qp is 160. If the memory holds the 'struct ibv_qp' was allocated at the upper boundary of a memory page, cast it to 'struct verbs_qp', whose size is 360, may across the memory page boundary. It will trigger invalid memory access to the next memory page. The issue can be reproduced with OPA and QIB HCA. For example run over OPA: Server Node: $ ib_read_bw -F -N -n 1000 -u 20 -q 257 -s 4194304 Client Node: $ ib_read_bw -F -N -n 1000 -u 20 -q 257 -s 4194304 <sever> Program received signal SIGSEGV, Segmentation fault. ibv_qp_to_qp_ex (qp=0x5555557a5f10) at libibverbs/verbs.c:624 624 if (vqp->comp_mask & VERBS_QP_EX) (gdb) bt #0 ibv_qp_to_qp_ex (qp=0x5555557a5f10) at libibverbs/verbs.c:624 #1 0x000055555556af4a in create_reg_qp_main (ctx=ctx@entry=0x7fffffffd500, user_param=user_param@entry=0x7fffffffd670, i=i@entry=21, num_of_qps=num_of_qps@entry=128) at src/perftest_resources.c:1597 #2 0x000055555556b6d7 in create_qp_main (num_of_qps=<optimized out>, i=21, user_param=0x7fffffffd670, ctx=0x7fffffffd500) at src/perftest_resources.c:1613 #3 ctx_init (ctx=0x7fffffffd500, user_param=0x7fffffffd670) at src/perftest_resources.c:1552 #4 0x0000555555558e9c in main (argc=<optimized out>, argv=<optimized out>) at src/read_bw.c:149 624 if (vqp->comp_mask & VERBS_QP_EX) (gdb) p qp $1 = (struct ibv_qp *) 0x5555557a5f10 (gdb) p vqp $2 = (struct verbs_qp *) 0x5555557a5f10 (gdb) p *qp $3 = {context = 0x55555578ad00, qp_context = 0x0, .... (gdb) p *vqp Cannot access memory at address 0x5555557a6000 Signed-off-by: Honggang Li <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
this branch is: