On Tue, Nov 28, 2017 at 07:04:33PM +0200, Amir Goldstein wrote: > On Tue, Nov 28, 2017 at 5:59 PM, Vivek Goyal <vgoyal@xxxxxxxxxx> wrote: > > Hi, > > > > Here is the V8 of the patches. Primarily I have taken care of comments > > from Amir. > > > > This applies on top of the fix I posted with subject "ovl: Pass > > ovl_get_nlink() parameters in right order" > > > > Amir, I have added one more patch int series where I pass index dentry > > to ovl_get_nlink() to get rid of a warning during testing. Amir, can you please have a look at patch 1 in the series. That's new patch as compared to V7. > > > > I think only outstanding concern with this patch series now is how to > > detect that metacopy feature was ever enabled and there are some > > metacopy only files present. In that case we probably don't want to > > allow user to mount with metacopy=off. We currently don't seem to have > > a generic infrastructure which keeps track of features and backward > > compatibility stuff etc. Discussion still seems to be on upstream. > > > > Is this a blocker for this patch series or this is something we can > > live with? Only downside of mounting a file system which has metacopied > > files (with metacopy=off), is that user will see truncated files. > > IMO it is fine to mount with metacopy=off, but user should not see > truncated file. IMO if metacopy=off, overlayfs still needs to copy up > data from origin if origin can be decoded and if origin cannot be decoded > user should get -EIO. Right. That's the current behavior with the patch. I forgot to mention that user will see truncated files only with old kernels. > Mounting an overlayfs with metacopies with kernel that does not support > metacopy will have unexpected results, same as with redirect_dir. Right. As of now old kernel should see truncated files. Vivek -- 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
