Skip to content

v3.1.x: UCX osc: properly release exclusive lock to avoid lockup #6972

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jsquyres merged 1 commit into from
Jan 6, 2020
Merged

v3.1.x: UCX osc: properly release exclusive lock to avoid lockup #6972

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 5 additions & 7 deletions ompi/mca/osc/ucx/osc_ucx_passive_target.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,23 +89,21 @@ static inline int start_exclusive(ompi_osc_ucx_module_t *module, int target) {
}

static inline int end_exclusive(ompi_osc_ucx_module_t *module, int target) {
uint64_t result_value = 0;
ucp_ep_h ep = OSC_UCX_GET_EP(module->comm, target);
ucp_rkey_h rkey = (module->state_info_array)[target].rkey;
uint64_t remote_addr = (module->state_info_array)[target].addr + OSC_UCX_STATE_LOCK_OFFSET;
ucs_status_t status;

status = ucp_atomic_swap64(ep, TARGET_LOCK_UNLOCKED,
remote_addr, rkey, &result_value);
if (status != UCS_OK) {
status = ucp_atomic_post(ep, UCP_ATOMIC_POST_OP_ADD,
-((int64_t)TARGET_LOCK_EXCLUSIVE), sizeof(uint64_t),
remote_addr, rkey);
if (UCS_OK != status) {
opal_output_verbose(1, ompi_osc_base_framework.framework_output,
"%s:%d: ucp_atomic_swap64 failed: %d\n",
"%s:%d: ucp_atomic_post failed: %d\n",
__FILE__, __LINE__, status);
return OMPI_ERROR;
}

assert(result_value >= TARGET_LOCK_EXCLUSIVE);

return OMPI_SUCCESS;
}

Expand Down