Hi, Please find attached V2 of patches. I think I have taken care of feedback from V1. Changes from V1 are as follows. - Made metacopy up feature dependent on index=on. - Added a patch to provide kernel config option, module option and mount option to enable metadata copyup feature. - Now I am copying up size from lower when metadata only copyup takes place. - I am returning -ESTALE in ovl_lookup() if an inode as METACOPY xattr but can find/resolve associated origin. - Calling ovl_set_attr() after metadata copyup so that mtime on lower and upper are same. I have little concern about access to METACOPY flag of inode stored in ovl_inode. We are accessing this flag in lockless manner from reader side and updating this flag under a mutex. My understanding is that current code will make sure that user will see this flag set only after METACOPY XATTR has been set (and not before that). Following is code flow. ovl_check_setxattr(OVL_XATTR_METACOPY) smp_wmb(); (From ovl_inode_update()) ovl_set_flag(OVL_METACOPY, d_inode(c->dentry)) But there is no lockless_deference() call while checking this flag in ovl_test_flag(). So I have two questions. Is this sufficient? And is this too subtle and prone to breakage. ------------------- In one of the recent converstions, people mentioned that chown/chmod lead to copy up files as well as data. We could optimize it so that only metadata is copied up during chown/chmod and data is copied up when file is opened for WRITE. This optimization potentially could be useful with containers and user namespaces. In popular scenario, people end up doing chown() on whole image directory tree based on container mappings. And this chown copies up everything, breaking sharing of page cache between containers. With these patches, only metadat is copied up during chown() and if file is opened for READ, d_real() returns lower dentry/inode. That way, different containers can still continue to use page cache. That's the use case I have in mind. Basically, I am relying on storing OVL_XATTR_ORIGIN in upper inode during copy up. I use that information to get to lower inode later and do data copy up when necessary. I also store OVL_XATTR_METACOPY in upper inode to mark that only metadata has been copied up and data copy up still might be required. Any feedback is helpful. Vivek Goyal (8): ovl: ovl_check_setxattr() get rid of redundant -EOPNOTSUPP check ovl: During copy up, first copy up metadata and then data ovl: Provide a mount option metacopy=on/off for metadata copyup ovl: Copy up only metadata during copy up where it makes sense ovl: Set xattr OVL_XATTR_METACOPY on upper file ovl: Set OVL_METACOPY flag during ovl_lookup() ovl: Return lower dentry if only metadata copy up took place ovl: Fix ovl_getattr() to get number of blocks from lower fs/overlayfs/Kconfig | 9 ++++ fs/overlayfs/copy_up.c | 115 ++++++++++++++++++++++++++++++++++++++--------- fs/overlayfs/inode.c | 21 +++++++-- fs/overlayfs/namei.c | 38 ++++++++++++++++ fs/overlayfs/overlayfs.h | 5 ++- fs/overlayfs/ovl_entry.h | 1 + fs/overlayfs/super.c | 36 +++++++++++++++ fs/overlayfs/util.c | 32 ++++++++----- 8 files changed, 219 insertions(+), 38 deletions(-) -- 2.13.5 -- 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
