[LU-17189] GDS is not working on master Created: 12/Oct/23  Updated: 25/Jan/24

Status: Open
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: Lustre 2.16.0

Type: Bug Priority: Critical
Reporter: Jinshan Xiong Assignee: WC Triage
Resolution: Unresolved Votes: 0
Labels: None

Issue Links:
Related
Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

GDS on Lustre is not working any more; it's broken by this commit f792297212387c2ff1b1dd6ead1e2d451a375ade around the same time last year.

In particular, this line and similar ones:
```
tx->tx_gpu = msg_md ? (msg_md->md_flags & LNET_MD_FLAG_GPU) : 0;
```
is problematic since `tx_gpu` is a single bit. It will take the last bit of `(msg_md->md_flags & LNET_MD_FLAG_GPU)` in this assignment that will definitely be zero.

I think it's fair to have a dedicated function to do this check. Patch is coming soon



 Comments   
Comment by Jinshan Xiong [ 12/Oct/23 ]

I don't have an account on review.whamcloud.com so I won't be able to create PR; but the fix should be pretty easy.

Comment by Peter Jones [ 13/Oct/23 ]

Jinshan

Nice to see you involved in Lustre again! If you want to reach out to me directly (or indirectly via info@whamcloud.com) then we can get you a working Gerrit account sorted out without too much bother...

Peter

Comment by Gerrit Updater [ 14/Oct/23 ]

"jinshan.xiong <jinshan.xiong@gmail.com>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/52702
Subject: LU-17189 o2ib: assign tx_gpu properly
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: cdd3049ec6a1ceb6675e2ed30a4a5430d1536d32

Generated at Sat Feb 10 03:33:22 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.