[LU-10213] o2iblnd: Potential discrepancy when allocating qp Created: 08/Nov/17  Updated: 24/Nov/20  Resolved: 28/Mar/18

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: Lustre 2.11.0, Lustre 2.10.7

Type: Bug Priority: Major
Reporter: Amir Shehata (Inactive) Assignee: Amir Shehata (Inactive)
Resolution: Fixed Votes: 0
Labels: None

Issue Links:
Duplicate
Related
is related to LU-7124 MLX5: Limit hit in cap.max_send_wr Resolved
is related to LU-12385 2.12.2: IOR read testing reliably ind... Resolved
is related to LU-12901 Failing to create a properly sized IB... Resolved
Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

patch https://review.whamcloud.com/#/c/18347 introduced a bug. Basically you're negotiating the number of fragments and then explicitly setting the max_send_wrs according to that number. This patch introduces the logic where if you fail to allocate the qp then you reduce max_send_wrs by quarter, but that could cause a problem where not enough wrs are available under heavy load, if you're sending multiple fragments.

 Comments   
Comment by Gerrit Updater [ 29/Nov/17 ]

Amir Shehata (amir.shehata@intel.com) uploaded a new patch: https://review.whamcloud.com/30310
Subject: LU-10213 lnd: calculate qp max_send_wrs properly
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 3e4336fbe8bfb166fff04562f2d2b6dfebd15661

Comment by Gerrit Updater [ 22/Dec/17 ]

Oleg Drokin (oleg.drokin@intel.com) merged in patch https://review.whamcloud.com/30310/
Subject: LU-10213 lnd: calculate qp max_send_wrs properly
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 017d328fa832697533e4e032fe9a9213ea105320

Comment by Joseph Gmitter (Inactive) [ 28/Mar/18 ]

Patch landed to master for 2.11.0

Comment by Gerrit Updater [ 07/Jan/19 ]

Minh Diep (mdiep@whamcloud.com) uploaded a new patch: https://review.whamcloud.com/33975
Subject: LU-10213 lnd: calculate qp max_send_wrs properly
Project: fs/lustre-release
Branch: b2_10
Current Patch Set: 1
Commit: 83ef8cc6ff97be2049b18df233b20740054c216d

Comment by Gerrit Updater [ 19/Jan/19 ]

Oleg Drokin (green@whamcloud.com) merged in patch https://review.whamcloud.com/33975/
Subject: LU-10213 lnd: calculate qp max_send_wrs properly
Project: fs/lustre-release
Branch: b2_10
Current Patch Set:
Commit: ce37c38691196075863eff6bb3ac9c6277e83f74

Comment by Patrick Farrell (Inactive) [ 16/Aug/19 ]

ashehata,

I think we need to revert this from 2.12.x and master before the 2.12.3 and master releases, respectively, to avoid the credits bug.

I tried to spin the patch, but I ran in to https://review.whamcloud.com/#/c/34945/, which depends on this and I don't really understand.

Would you be able to generate appropriate revert patches for those branches?

Generated at Sat Feb 10 02:33:03 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.