Re: 9p EBADF with cache enabled (Was: 9p fs-cache tests/benchmark (was: 9p fscache Duplicate cookie detected))

From: Christian Schoenebeck
Date: Tue Apr 26 2022 - 11:38:48 EST

On Montag, 25. April 2022 16:10:16 CEST David Howells wrote:
> There may be a quick and dirty workaround. I think the problem is that
> unless the O_APPEND read starts at the beginning of a page, netfs is going
> to enforce a read. Does the attached patch fix the problem? (note that
> it's untested)

Patch doesn't apply for me on master:

checking file fs/9p/vfs_addr.c
Hunk #1 FAILED at 291.
1 out of 1 hunk FAILED
checking file fs/netfs/buffered_read.c
Hunk #1 FAILED at 364.
1 out of 1 hunk FAILED
checking file fs/netfs/internal.h
checking file fs/netfs/stats.c
Hunk #2 FAILED at 38.
1 out of 2 hunks FAILED

commit d615b5416f8a1afeb82d13b238f8152c572d59c0 (HEAD -> master, origin/master, origin/HEAD)
Merge: 0fc74d820a01 4d8ec9120819
Author: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Date: Mon Apr 25 10:53:56 2022 -0700

What was is based on?

> Also, can you get the contents of /proc/fs/fscache/stats from after
> reproducing the problem?

FS-Cache statistics
Cookies: n=684 v=1 vcol=0 voom=0
Acquire: n=689 ok=689 oom=0
LRU : n=0 exp=0 rmv=0 drp=0 at=0
Invals : n=0
Updates: n=2095 rsz=0 rsn=0
Relinqs: n=5 rtr=0 drop=5
NoSpace: nwr=0 ncr=0 cull=0
IO : rd=0 wr=0
RdHelp : RA=974 RP=0 WB=13323 WBZ=2072 rr=0 sr=0
RdHelp : ZR=13854 sh=0 sk=2072
RdHelp : DL=14297 ds=14297 df=13322 di=0
RdHelp : RD=0 rs=0 rf=0
RdHelp : WR=0 ws=0 wf=0

Best regards,
Christian Schoenebeck