Skip to content

Patches needed for Stable

Boaz Harrosh linux-nfs 1 version New

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest6 patches 0 repliesNew

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/6pnfs: save layoutcommit lwb at layout headerfs/nfs/nfs4filelayout.c fs/nfs/pnfs.c fs/nfs/pnfs.h+8 -60 replies
2/6pnfs: save layoutcommit cred at layout header initfs/nfs/pnfs.c fs/nfs/pnfs.h+12 -110 replies
3/6pnfs: let layoutcommit handle a list of lsegfs/nfs/nfs4proc.c fs/nfs/pnfs.c fs/nfs/pnfs.h +1 more+26 -180 replies
4/6pnfs: use lwb as layoutcommit lengthfs/nfs/nfs4xdr.c+1 -10 replies
5/6pnfs-obj: Bug when we are running out of biofs/nfs/objlayout/objio_osd.c+5 -70 replies
6/6pnfs-obj: Fix the comp_index != 0 casefs/nfs/objlayout/objio_osd.c fs/nfs/objlayout/pnfs_osd_xdr_cli.c+10 -90 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [PATCHSET 0/6] Patches needed for Stable

Trond Hi

I've played all day with these commit patches and I came to the conclusion that we will need all 5 of Peng's commit patches. The minimal set is the first 4 and without the 5th one, is as you pointed out, might be very racy. Let me explain.

At first I thought all I needed was:
[PATCH v4 05/27] pnfs: let layoutcommit handle a list of lseg [PATCH v4 06/27] pnfs: use lwb as layoutcommit length

But They do not apply as is, and would not work.
I tried to, instead of taking the creds and lwb from the *header*, take them from the *first segment*. But this is an ugly hack (I tried, it's really ugly) and is scary as hell. Because it is code we never tested as opposed to this code that was the only one I actually tested since before last bakeathon. I would hate to put such under-tested code in Stable when we will change it hopefully in this Kernel and is ugly as hell.

So Introducing these two:
[PATCH v4 03/27] pnfs: save layoutcommit lwb at layout header [PATCH v4 04/27] pnfs: save layoutcommit cred at layout header
Just makes sense.

Now also:
[PATCH v4 07/27] NFS41: save layoutcommit cred in layout header init

Was pointed by you as fixing a potential race, which I agree completely, and I think it should be just merged with:
[PATCH 2/6] pnfs: save layoutcommit cred at layout header
Which are the same exact issue only done better

I'm adding two more patches I'll need for Stable which are bug fixes we found in the large testing we did here after Bakeathon:

	[PATCH 5/6] pnfs-obj: Bug when we are running out of bio
	[PATCH 6/6] pnfs-obj: Fix the comp_index != 0 case

----
So here is the list of patches for stable:

[PATCH 1/6] pnfs: save layoutcommit lwb at layout header [PATCH 2/6] pnfs: save layoutcommit cred at layout header init [PATCH 3/6] pnfs: let layoutcommit handle a list of lseg [PATCH 4/6] pnfs: use lwb as layoutcommit length

These 4 are Peng's 5 layoutcommit patches. I've combined
[PATCH v4 04/27] pnfs: save layoutcommit cred at layout header
and
[PATCH v4 07/27] NFS41: save layoutcommit cred in layout header init
Which are just the same issue. And rebased the last one on top of that. And added CC: Stable.

[PATCH 5/6] pnfs-obj: Bug when we are running out of bio [PATCH 6/6] pnfs-obj: Fix the comp_index != 0 case

Thanks
Boaz