Re: [dhowells-fs:cifs-netfs] [cifs] b4834f12a4: WARNING:at_fs/netfs/write_collect.c:#netfs_writeback_lookup_folio

[Oliver Sang <oliver.sang@intel.com>]

hi, David,

we actually hope to seek your help as below.

On Fri, Apr 26, 2024 at 10:42:41AM +0100, David Howells wrote:
> Trying to do "lkp run job.yaml" doesn't work:
> 
> /root/lkp-tests/filters/need_kconfig.rb:20:in `read_kernel_kconfigs': .config doesn't exist: /pkg/linux/x86_64-rhel-8.3/gcc-13/b4834f12a4df607aaedc627fa9b93f3b18f664ba/.config (Job::ParamError)
>         from /root/lkp-tests/filters/need_kconfig.rb:176:in `block in expand_expression'
>         from /root/lkp-tests/lib/erb.rb:51:in `eval'
>         from /root/lkp-tests/lib/erb.rb:51:in `expand_expression'
>         from /root/lkp-tests/lib/job.rb:646:in `evaluate_param'
>         from /root/lkp-tests/lib/job.rb:694:in `block in expand_params'
>         from /root/lkp-tests/lib/job.rb:79:in `block in for_each_in'
>         from /root/lkp-tests/lib/job.rb:78:in `each'
>         from /root/lkp-tests/lib/job.rb:78:in `for_each_in'
>         from /root/lkp-tests/lib/job.rb:691:in `expand_params'
>         from /root/lkp-tests/bin/run-local:138:in `<main>'
> 
> I tried to run the filebench directly, but that only wants to hammer on
> /tmp/bigfileset/ and also wants a file for SHM precreating in /tmp.  I was
> able to get it to work with cifs by:
> 
> touch /tmp/filebench-shm-IF6uX8
> truncate -s 184975240 /tmp/filebench-shm-IF6uX8
> mkdir /tmp/bigfileset
> mount //myserver/test /tmp/bigfileset/ -o user=shares,pass=...,cache=loose
> 
> /root/lkp-tests/programs/filebench/pkg/filebench-lkp/lkp/benchmarks/filebench/bin/filebench -f /lkp/benchmarks/filebench/share/filebench/workloads/filemicro_seqwriterandvargam.f
> 
> It tries to remove /tmp/bigfileset/, can't because it's mounted, and then
> continues anyway.
> 
> It should be easier than this ;-)

really sorry about so many troubles.

we made some clean and fix recently, then the reproducer can run quite well on
debian now.

however, same process still blocks on fedora-39 side, while

mount -t cifs -o user=root,password=pass //localhost/fs/sda3 /cifs/sda3

seems give us a right result:

//localhost/fs/sda3 on /cifs/sda3 type cifs (rw,relatime,vers=3.1.1,cache=strict,username=root,uid=0,noforceuid,gid=0,noforcegid,addr=0000:0000:0000:0000:0000:0000:0000:0001,file_mode=0755,dir_mode=0755,soft,nounix,serverino,mapposix,rsize=4194304,wsize=4194304,bsize=1048576,echo_interval=60,actimeo=1,closetimeo=1)

the following steps directly cause core dump (which runs well on debian)

2024-05-14 00:40:11 sync
2024-05-14 00:40:11 echo 3 > /proc/sys/vm/drop_caches
/home/sdp/lkp-tests/lib/job.sh: line 229: 41437 Bus error               (core dumped) "$@"
...


we are still checking now. just was wondering if you would have some quick
guidance about cifs mount to us what could cause this?


BTW, we noticed you sent a patch for this report:
https://lore.kernel.org/lkml/2150448.1714130115@warthog.procyon.org.uk/

and we also noticed this patch is already merged into latest
  branch: dhowells-fs/cifs-netfs

by our tests, the WARNING in this report disappeared in latest branch, however
we noticed some other issues which we still check if our bisect is correct.


> 
> David
>