Rohith Surabattula <rohiths.msft@xxxxxxxxx> wrote:
> I noticed 2 other issues while running xfstests.
>
> Noticed kernel OOPS during test generic/286:
> folio_test_writeback returned false which means PG_writeback flag has
> been cleared. I am not sure whether head page has PG_writeback flag
> set initially? Can you please confirm.
>
> [ 2275.941096] CIFS: bad 2000 @64f0000 page 64f0 64f1
> [ 2275.945785] ------------[ cut here ]------------
> [ 2275.945787] kernel BUG at /home/lxsmbadmin/latest_14mar/linux-fs/fs/cifs/cifssmb.c:1954!
> ...
> [ 2275.969812] Workqueue: cifsiod cifs_writev_complete [cifs]
> [ 2275.974909] RIP: 0010:cifs_pages_written_back+0x1e1/0x1f0 [cifs]
> [ 2275.975570] CIFS: bad 2000 @64f0000 page 64f0 64f1
> [ 2275.980641] Call Trace:
> [ 2275.980643] <TASK>
> [ 2275.980648] cifs_writev_complete+0x43d/0x500 [cifs]
I don't see that, but it fails in some other ways. I think the bits should be
set.
I am seeing the occasional:
CIFS: trying to dequeue a deleted mid
but I haven't managed to work out how I get to that yet.
I'm also occasionally seeing cifs_open() return a number >0, which causes all
sorts of fun.
> Noticed that with netfs integration, file open with O_DIRECT flag is
> not supported.
It should be. It jumps off to netfs_direct_read_iter() in various places.
openat(AT_FDCWD, "/xfstest.test/hello2", O_RDWR|O_DIRECT) = 3
fstatfs(3, {f_type=SMB2_MAGIC_NUMBER, f_bsize=1024, f_blocks=54961016, f_bfree=23691924, f_bavail=23691924, f_files=0, f_ffree=0, f_fsid={val=[1904023890, 0]}, f_namelen=255, f_frsize=1024, f_flags=ST_VALID|ST_RELATIME}) = 0
fstat(3, {st_mode=S_IFREG|0755, st_size=278528, ...}) = 0
pread64(3, "\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253\253"..., 4096, 0) = 4096
Btw, I've pushed an update to my cifs-experimental branch.
David
