Unh, it seems i_writecount not work.
If we modify file through lowerfs, i_writecount won't change, but the
size already changed.
For example:
echo "111" > /lowerfs/test
ls -l /upper/test
echo "2222" >> /lowerfs/test
ls -l /upper/test
So, can you describe your test enviroment? including kernel version and
fsx parameters, I will check it.
Thanks.
On 2021/6/22 15:59, Miklos Szeredi wrote:
On Sat, 30 Jan 2021 at 09:50, Fengnan Chang <changfengnan@xxxxxxxx> wrote:
When writeback cache is enabled, the inode information in cached is
considered new by default, and the inode information of lowerfs is
stale.
When a lower fs is mount in a different directory through different
connection, for example PATHA and PATHB, since writeback cache is
enabled by default, when the file is modified through PATHA, viewing the
same file from the PATHB, PATHB will think that cached inode is newer
than lowerfs, resulting in file size and time from under PATHA and PATHB
is inconsistent.
Add a judgment condition to check whether to use the info in the cache
according to mtime.
This seems to break the fsx-linux stress test.
I suspect a better direction would be looking at whether the inode has
any files open for write (i_writecount > 0)...
Thanks,
Miklos
