On Wed, Oct 18, 2017 at 12:05 AM, Vivek Goyal <vgoyal@xxxxxxxxxx> wrote:
> OVL_METACOPY in oi->flags can be accessed in lockless manner. So if a file
> is being copied up metadata only, we need to make sure that upperdentry is
> visible only after OVL_METACOPY flag has been set. IOW, if oi->__upperdentry
> is visble to a cpu, then we also need to make sure any updates to OVL_METACOPY
> flags are visible too.
>
You know, I have a feeling that this ordering requirement could be simplified or
completely avoided if you flip the meaning of the flag, i.e.:
bool ovl_dentry_has_upper_data(struct dentry *dentry)
{
return ovl_test_flag(OVL_UPPER_DATA, d_inode(dentry));
}
Then flag visibility requirements are the same as visibility requirements
for oe->has_upper.
You probably don't need to add any new barriers for setting setting the flag
on normal copy up.
For setting the flag in copy_up_meta_inode_data, and testing the flag
in ovl_d_real() the requirements stay the same as you implemented in patch 10.
Am I wrong?
--
To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
