[LU-12780] Avoid using ptlrpc_thread where is in't needed

• Type: Improvement
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.15.0
• Assignee: Neil Brown
• Reporter: Neil Brown

Description

Several places in lustre use 'struct ptlrpc_thread' to "help" manage a kthread.

These places are not directly related to ptlrpc, so the naming is not help, and in general the usage of ptlrpc_thread does not improve the code.

By removing the use of ptlrpc_thread - particularly by avoiding startup synchornization and by using kthread_stop/ kthread_should_stop(), I can reduce code size by a couple of hundred lines.

Using the seemly uniform approach of ptlrpc_thread can lead to an impression that these threads are all managed much the same way, but this is not the case.  There is substantial variety on how the threads work and exposing that clear will benefit maintainability.

 

Issue Links

• is related to: LU-13986 livelock is possible in distribute_txn_commit_thread()
• is related to: LU-12511 Prepare lustre for adoption into the linux kernel

Comments

[Peter Jones]

Landed for 2.15

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/36262/
Subject: LU-12780 ofd: don't use ptlrpc_thread for consistency verification
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: cb3c65d4a10b2ad0a6c70d5e38719bca3d501efb

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/38824/
Subject: LU-12780 osd: use native kthreads for scrub.
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: b76df292c241a6db8dfad76d0f5f4cb8babe22bf

[Neil Brown]

(oops, I meant to reply to the above, but inadvertently editted the comment instead.  I didn't know I could do that, but think I've put it back properly).

 

> Are there any known bugs with scrub that are know to be fixed or might be fixed by the change?

I didn't find any clear races, but I didn't look very hard.  It is probably that the flags are changed very rarely, so any races that are possible are extremely unlikely.

My main reason for the patch was that I was going to add another flag which could change asynchronously, and I didn't think that would be safe without proper locking.

 

[Cory Spitz]

From https://review.whamcloud.com/38974:

If two of these updates can race, corruption can occurs

Are there any known bugs with scrub that are know to be fixed or might be fixed by the change?

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/38974/
Subject: LU-12780 scrub: all update to bitfields must be protected.
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 23a6e1ed8eec2c07653ed07c35bb109ecb87a5b7

[Gerrit Updater]

Neil Brown ([e-mail]) uploaded a new patch: https://review.whamcloud.com/38974
Subject: LU-12780 scrub: all update to bitfields must be protected.
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f8ff777659f2152528085c77789c08fcca81301d

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/38612/
Subject: LU-12780 quota: don't use ptlrpc_thead of qmt_pool_recalc
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: d9094cf66b54c953416969e8581e1f3a6df461b0

[Gerrit Updater]

Neil Brown ([e-mail]) uploaded a new patch: https://review.whamcloud.com/38824
Subject: LU-12780 osd: use native kthreads for scrub.
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 562bf522f156df53d4d298fd9b144398301b50d8

[Gerrit Updater]

Neil Brown ([e-mail]) uploaded a new patch: https://review.whamcloud.com/38612
Subject: LU-12780 quota: don't use ptlrpc_thead of qmt_pool_recalc
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 6fa73578014258098e0f86e8b448d8fbb2290a8d