Bug 274715 - [ktls] deadlock with mlx5 and KTLS receive
Summary: [ktls] deadlock with mlx5 and KTLS receive
Status: Closed FIXED
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: CURRENT
Hardware: Any Any
: --- Affects Only Me
Assignee: freebsd-net (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-10-25 17:48 UTC by Martin Matuska
Modified: 2023-11-22 08:35 UTC (History)
1 user (show)

See Also:

Attachments
DDB thread 1 (186.47 KB, image/jpeg)
2023-10-25 17:49 UTC, Martin Matuska 		no flags Details
DDB thread 2 (193.40 KB, image/jpeg)
2023-10-25 17:50 UTC, Martin Matuska 		no flags Details
DDB lockchain (194.61 KB, image/jpeg)
2023-10-25 17:50 UTC, Martin Matuska 		no flags Details
DDB trace (text) (3.57 KB, text/plain)
2023-10-25 18:09 UTC, Martin Matuska 		no flags Details
DDB trace (text) (3.65 KB, text/plain)
2023-10-25 18:12 UTC, Martin Matuska 		no flags Details
Show Obsolete (1) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Martin Matuska freebsd_committer freebsd_triage 2023-10-25 17:48:24 UTC 

    


    


      



        
          Comment 1
        

        
          Martin Matuska

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-10-25 17:49:51 UTC
        

      






Created attachment 245864 [details]
DDB thread 1

    


    


      



        
          Comment 2
        

        
          Martin Matuska

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-10-25 17:50:08 UTC
        

      






Created attachment 245865 [details]
DDB thread 2

    


    


      



        
          Comment 3
        

        
          Martin Matuska

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-10-25 17:50:39 UTC
        

      






Created attachment 245866 [details]
DDB lockchain

    


    


      



        
          Comment 4
        

        
          Martin Matuska

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-10-25 18:09:42 UTC
        

      






Created attachment 245867 [details]
DDB trace (text)

    


    


      



        
          Comment 5
        

        
          Martin Matuska

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-10-25 18:12:23 UTC
        

      






Created attachment 245868 [details]
DDB trace (text)

    


    


      



        
          Comment 6
        

        
          Konstantin Belousov

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-10-27 22:47:40 UTC
        

      






So who is the owner of the lock blocking other threads?

The min useful set of data for a deadlock is ddb 'show alllocks' and 'alltrace'.

    


    


      



        
          Comment 7
        

        
          Martin Matuska

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-10-28 10:18:47 UTC
        

      






I have written a patch for review:
https://reviews.freebsd.org/D42368

    


    


      



        
          Comment 8
        

        
          commit-hook

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-11-16 11:17:54 UTC
        

      






A commit in branch main references this bug:

URL: https://cgit.FreeBSD.org/src/commit/?id=a592812327deaf69ab226afc5c8a01af43dc03c2

commit a592812327deaf69ab226afc5c8a01af43dc03c2
Author:     Martin Matuska <mm@FreeBSD.org>
AuthorDate: 2023-11-13 13:29:27 +0000
Commit:     Martin Matuska <mm@FreeBSD.org>
CommitDate: 2023-11-16 11:17:41 +0000

    mlx5_core: fix deadlock when using RXTLS

    If removing a node of type FS_TYPE_FLOW_DEST we lock the flow group too
    late. This can lead to a deadlock with fs_add_dst_fg().

    PR:             274715
    MFC after:      1 week
    Reviewed by:    kib
    Tested by:      mm
    Differential Revision: https://reviews.freebsd.org/D42368

 sys/dev/mlx5/mlx5_core/mlx5_fs_tree.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

    


    


      



        
          Comment 9
        

        
          commit-hook

        

        
            freebsd_committer
            freebsd_triage
        

        
          2023-11-22 08:01:24 UTC
        

      






A commit in branch stable/14 references this bug:

URL: https://cgit.FreeBSD.org/src/commit/?id=ac83e415ef083da400509c6fa434ee59cfe9668f

commit ac83e415ef083da400509c6fa434ee59cfe9668f
Author:     Martin Matuska <mm@FreeBSD.org>
AuthorDate: 2023-11-13 13:29:27 +0000
Commit:     Martin Matuska <mm@FreeBSD.org>
CommitDate: 2023-11-22 08:00:23 +0000

    mlx5_core: fix deadlock when using RXTLS

    If removing a node of type FS_TYPE_FLOW_DEST we lock the flow group too
    late. This can lead to a deadlock with fs_add_dst_fg().

    PR:             274715
    Reviewed by:    kib
    Tested by:      mm
    Differential Revision: https://reviews.freebsd.org/D42368

    (cherry picked from commit a592812327deaf69ab226afc5c8a01af43dc03c2)

 sys/dev/mlx5/mlx5_core/mlx5_fs_tree.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)