-
Notifications
You must be signed in to change notification settings - Fork 0
Conversation
Create bee file with cp linux-5.10.113-434.bee linux-5.15.69-435.bee and update SRCURL.
Create bee file with cp nvidia_linux-5.10.113-434-510.60.02-0.bee nvidia_linux-5.15.69-435-510.60.02-0.bee
later a process was hanging with
|
And again (from a new boot)
and yet again
|
I can not tell how helpfull it is for the current case. But I had no NFS issues whilst trying an unpatched (nfstop and friends) 5.15.56 kernel on otherland (see messages.2022-08 ibd.). |
Use mariux-5.15.69-436 which added upstream commit 6e176d47160c ("NFSv4: Fixes for nfs4_inode_return_delegation()"): Author: Trond Myklebust <trond.myklebust@hammerspace.com> Date: Sun Oct 10 10:58:12 2021 +0200 NFSv4: Fixes for nfs4_inode_return_delegation() We mustn't call nfs_wb_all() on anything other than a regular file. Furthermore, we can exit early when we don't hold a delegation. Reported-by: David Wysochanski <dwysocha@redhat.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 11118398f495..7c9eb679dbdb 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -755,11 +755,13 @@ int nfs4_inode_return_delegation(struct inode *inode) struct nfs_delegation *delegation; delegation = nfs_start_delegation_return(nfsi); - /* Synchronous recall of any application leases */ - break_lease(inode, O_WRONLY | O_RDWR); - nfs_wb_all(inode); - if (delegation != NULL) + if (delegation != NULL) { + /* Synchronous recall of any application leases */ + break_lease(inode, O_WRONLY | O_RDWR); + if (S_ISREG(inode->i_mode)) + nfs_wb_all(inode); return nfs_end_delegation_return(inode, delegation, 1); + } return 0; }
@pmenzel got a patch from Trond. First test looked good. I've added that patch to https://github.molgen.mpg.de/mariux64/linux/commits/mariux-5.15.69-436 and will make a new bee package.
|
Installation instruction:
Pro-Tip: make sure, you are on the system you think you are. :-) |
There is no more /proc/[PID]/io file in the procfs :( Most probably caused by 'CONFIG_TASKSTATS is not set', what in turn disables 'TASK_IO_ACCOUNTING'. Is there a replacement available? |
Well, sorry for bothering. I still don't get it. In the 5.10 kernel the overlay-fs is available, what exactly is the new requirement? |
Indeed. Don't know why docker didn't enable its overlay2 driver on dose with the 5.10.113 kernel. |
Hmmm. Bug in build system? I don't get it yet.
|
TASKSTATS was auto.selected by KVM until mariux64/linux@63b3f96 which landed in v5.14. So we had TASKSTATS although its default is "N". |
Use mariux-5.15.69-437 which includes fix to reenable /proc/pid/io [1]. [1]: mariux64/linux@8d0e75ece55
615f5f8
to
1c50111
Compare
Rebuild with a new patch to supress sporadic ext4 warnings when mxqd uses fcntl EXT4_IOC_SHUTDOWN on its job tmpdirs. Update minor Linux version while we are at it.
Looks good to me.
You striked this through. Not needed in the end? |
Its not a 5.15 feature, its available in 5.10 as well. |
79da577
to
f745b0b
Compare
Latest kernel of latest LTR
Includes overlayfs (for e.g. rootless docker)Tested on theinternet (with nvidia)