8000 [TE FIX] Updating the outstanding work request counting when closing a QP by alogfans · Pull Request #390 · kvcache-ai/Mooncake · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

[TE FIX] Updating the outstanding work request counting when closing a QP #390

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
merged 2 commits into from
May 22, 2025
Merged

[TE FIX] Updating the outstanding work request counting when closing a QP #390

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d509073
[TE] When calling disconnectUnlocked and deconstruct, fix the countin…
alogfans May 22, 2025
52b35a3
reset wr_depth_list_ after deducting cq_outstanding
alogfans May 22, 2025
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
31 changes: 22 additions & 9 deletions mooncake-transfer-engine/src/transport/rdma_transport/rdma_endpoint.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,14 +77,21 @@ int RdmaEndPoint::construct(ibv_cq *cq, size_t num_qp_list,

int RdmaEndPoint::deconstruct() {
for (size_t i = 0; i < qp_list_.size(); ++i) {
if (wr_depth_list_[i] != 0)
LOG(WARNING)
<< "Outstanding work requests found, CQ will not be 8000 generated";

if (ibv_destroy_qp(qp_list_[i])) {
PLOG(ERROR) << "Failed to destroy QP";
return ERR_ENDPOINT;
}
// After destroying QP, the wr_depth_list_ won't change
bool displayed = false;
if (wr_depth_list_[i] != 0) {
if (!displayed) {
LOG(WARNING)
<< "Outstanding work requests found, CQ will not be generated";
displayed = true;
}
__sync_fetch_and_sub(cq_outstanding_, wr_depth_list_[i]);
wr_depth_list_[i] = 0;
}
}
qp_list_.clear();
delete[] wr_depth_list_;
Expand Down Expand Up @@ -218,18 +225,24 @@ void RdmaEndPoint::disconnect() {
}

void RdmaEndPoint::disconnectUnlocked() {
for (size_t i = 0; i < qp_list_.size(); ++i) {
if (wr_depth_list_[i] != 0)
LOG(WARNING) << "Outstanding work requests will be dropped";
}
ibv_qp_attr attr;
memset(&attr, 0, sizeof(attr));
attr.qp_state = IBV_QPS_RESET;
for (size_t i = 0; i < qp_list_.size(); ++i) {
int ret = ibv_modify_qp(qp_list_[i], &attr, IBV_QP_STATE);
if (ret) PLOG(ERROR) << "Failed to modify QP to RESET";
// After resetting QP, the wr_depth_list_ won't change
bool displayed = false;
if (wr_depth_list_[i] != 0) {
if (!displayed) {
LOG(WARNING)
<< "Outstanding work requests found, CQ will not be generated";
displayed = true;
}
__sync_fetch_and_sub(cq_outstanding_, wr_depth_list_[i]);
wr_depth_list_[i] = 0;
}
}
for (size_t i = 0; i < qp_list_.size(); ++i) wr_depth_list_[i] = 0;
status_.store(UNCONNECTED, std::memory_order_release);
}

Expand Down
Loading
0