Message ID | 1506585198-32494-2-git-send-email-nborisov@suse.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Thu, Sep 28, 2017 at 10:53:18AM +0300, Nikolay Borisov wrote: > Currently when a read-only snapshot is received and subsequently its ro property > is set to false i.e. switched to rw-mode the received_uuid of that subvol remains > intact. However, once the received volume is switched to RW mode we cannot > guaranteee that it contains the same data, so it makes sense to remove the > received uuid. The presence of the received_uuid can also cause problems when > the volume is being send. > > Signed-off-by: Nikolay Borisov <nborisov@suse.com> > --- > > v3: > * Rework the patch considering latest feedback from David Sterba i.e. > explicitly use btrfs_end_transaction > > fs/btrfs/ioctl.c | 36 +++++++++++++++++++++++++++++------- > 1 file changed, 29 insertions(+), 7 deletions(-) > > diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c > index ee4ee7cbba72..c0374125cec2 100644 > --- a/fs/btrfs/ioctl.c > +++ b/fs/btrfs/ioctl.c > @@ -1811,6 +1811,17 @@ static noinline int btrfs_ioctl_subvol_setflags(struct file *file, > goto out_drop_sem; > > root_flags = btrfs_root_flags(&root->root_item); > + > + /* > + * 1 - root item > + * 1 - uuid item > + */ > + trans = btrfs_start_transaction(root, 2); > + if (IS_ERR(trans)) { > + ret = PTR_ERR(trans); > + goto out_drop_sem; > + } > + > if (flags & BTRFS_SUBVOL_RDONLY) { > btrfs_set_root_flags(&root->root_item, > root_flags | BTRFS_ROOT_SUBVOL_RDONLY); > @@ -1824,22 +1835,33 @@ static noinline int btrfs_ioctl_subvol_setflags(struct file *file, > btrfs_set_root_flags(&root->root_item, > root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); > spin_unlock(&root->root_item_lock); > + if (!btrfs_is_empty_uuid(root->root_item.received_uuid)) { > + ret = btrfs_uuid_tree_rem(trans, fs_info, > + root->root_item.received_uuid, > + BTRFS_UUID_KEY_RECEIVED_SUBVOL, > + root->root_key.objectid); > + > + if (ret && ret != -ENOENT) { > + btrfs_abort_transaction(trans, ret); > + btrfs_end_transaction(trans); > + goto out_reset; > + } > + > + memset(root->root_item.received_uuid, 0, > + BTRFS_UUID_SIZE); > + } > } else { > spin_unlock(&root->root_item_lock); > btrfs_warn(fs_info, > "Attempt to set subvolume %llu read-write during send", > root->root_key.objectid); > ret = -EPERM; > - goto out_drop_sem; > + btrfs_abort_transaction(trans, ret); > + btrfs_end_transaction(trans); > + goto out_reset; Adding the transaction before the "if (flags & BTRFS_SUBVOL_RDONLY)" condition makes it much worse. The "is subvolume in send" test is supposed to be lightweight and should not shoot down the whole filesystem. The usecase is explained in 2c68653787f91c62f8. Also the received_uuid must be changed under the root_item_lock. I think it should be fine to keep the transaction start where it is, change the received_uuid eventually and let it commit. You can set the transaction units to 2 unconditionally. > } > } > > - trans = btrfs_start_transaction(root, 1); > - if (IS_ERR(trans)) { > - ret = PTR_ERR(trans); > - goto out_reset; > - } > - > ret = btrfs_update_root(trans, fs_info->tree_root, > &root->root_key, &root->root_item); > if (ret < 0) { -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 29.09.2017 20:56, David Sterba wrote: > On Thu, Sep 28, 2017 at 10:53:18AM +0300, Nikolay Borisov wrote: >> Currently when a read-only snapshot is received and subsequently its ro property >> is set to false i.e. switched to rw-mode the received_uuid of that subvol remains >> intact. However, once the received volume is switched to RW mode we cannot >> guaranteee that it contains the same data, so it makes sense to remove the >> received uuid. The presence of the received_uuid can also cause problems when >> the volume is being send. >> >> Signed-off-by: Nikolay Borisov <nborisov@suse.com> >> --- >> >> v3: >> * Rework the patch considering latest feedback from David Sterba i.e. >> explicitly use btrfs_end_transaction >> >> fs/btrfs/ioctl.c | 36 +++++++++++++++++++++++++++++------- >> 1 file changed, 29 insertions(+), 7 deletions(-) >> >> diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c >> index ee4ee7cbba72..c0374125cec2 100644 >> --- a/fs/btrfs/ioctl.c >> +++ b/fs/btrfs/ioctl.c >> @@ -1811,6 +1811,17 @@ static noinline int btrfs_ioctl_subvol_setflags(struct file *file, >> goto out_drop_sem; >> >> root_flags = btrfs_root_flags(&root->root_item); >> + >> + /* >> + * 1 - root item >> + * 1 - uuid item >> + */ >> + trans = btrfs_start_transaction(root, 2); >> + if (IS_ERR(trans)) { >> + ret = PTR_ERR(trans); >> + goto out_drop_sem; >> + } >> + >> if (flags & BTRFS_SUBVOL_RDONLY) { >> btrfs_set_root_flags(&root->root_item, >> root_flags | BTRFS_ROOT_SUBVOL_RDONLY); >> @@ -1824,22 +1835,33 @@ static noinline int btrfs_ioctl_subvol_setflags(struct file *file, >> btrfs_set_root_flags(&root->root_item, >> root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); >> spin_unlock(&root->root_item_lock); >> + if (!btrfs_is_empty_uuid(root->root_item.received_uuid)) { >> + ret = btrfs_uuid_tree_rem(trans, fs_info, >> + root->root_item.received_uuid, >> + BTRFS_UUID_KEY_RECEIVED_SUBVOL, >> + root->root_key.objectid); >> + >> + if (ret && ret != -ENOENT) { >> + btrfs_abort_transaction(trans, ret); >> + btrfs_end_transaction(trans); >> + goto out_reset; >> + } >> + >> + memset(root->root_item.received_uuid, 0, >> + BTRFS_UUID_SIZE); >> + } >> } else { >> spin_unlock(&root->root_item_lock); >> btrfs_warn(fs_info, >> "Attempt to set subvolume %llu read-write during send", >> root->root_key.objectid); >> ret = -EPERM; >> - goto out_drop_sem; >> + btrfs_abort_transaction(trans, ret); >> + btrfs_end_transaction(trans); >> + goto out_reset; > > Adding the transaction before the "if (flags & BTRFS_SUBVOL_RDONLY)" > condition makes it much worse. The "is subvolume in send" test is > supposed to be lightweight and should not shoot down the whole > filesystem. The usecase is explained in 2c68653787f91c62f8. > > Also the received_uuid must be changed under the root_item_lock. > > I think it should be fine to keep the transaction start where it is, > change the received_uuid eventually and let it commit. You can set the > transaction units to 2 unconditionally. So what you are suggesting is to not move the transaction start before the if check? But then how would you structure the code to remove the uuid only if we are switchin RO->RW and not in send without duplicating the checks right before btrfs_update_root? > >> } >> } >> >> - trans = btrfs_start_transaction(root, 1); >> - if (IS_ERR(trans)) { >> - ret = PTR_ERR(trans); >> - goto out_reset; >> - } >> - >> ret = btrfs_update_root(trans, fs_info->tree_root, >> &root->root_key, &root->root_item); >> if (ret < 0) { -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index ee4ee7cbba72..c0374125cec2 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -1811,6 +1811,17 @@ static noinline int btrfs_ioctl_subvol_setflags(struct file *file, goto out_drop_sem; root_flags = btrfs_root_flags(&root->root_item); + + /* + * 1 - root item + * 1 - uuid item + */ + trans = btrfs_start_transaction(root, 2); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + goto out_drop_sem; + } + if (flags & BTRFS_SUBVOL_RDONLY) { btrfs_set_root_flags(&root->root_item, root_flags | BTRFS_ROOT_SUBVOL_RDONLY); @@ -1824,22 +1835,33 @@ static noinline int btrfs_ioctl_subvol_setflags(struct file *file, btrfs_set_root_flags(&root->root_item, root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); spin_unlock(&root->root_item_lock); + if (!btrfs_is_empty_uuid(root->root_item.received_uuid)) { + ret = btrfs_uuid_tree_rem(trans, fs_info, + root->root_item.received_uuid, + BTRFS_UUID_KEY_RECEIVED_SUBVOL, + root->root_key.objectid); + + if (ret && ret != -ENOENT) { + btrfs_abort_transaction(trans, ret); + btrfs_end_transaction(trans); + goto out_reset; + } + + memset(root->root_item.received_uuid, 0, + BTRFS_UUID_SIZE); + } } else { spin_unlock(&root->root_item_lock); btrfs_warn(fs_info, "Attempt to set subvolume %llu read-write during send", root->root_key.objectid); ret = -EPERM; - goto out_drop_sem; + btrfs_abort_transaction(trans, ret); + btrfs_end_transaction(trans); + goto out_reset; } } - trans = btrfs_start_transaction(root, 1); - if (IS_ERR(trans)) { - ret = PTR_ERR(trans); - goto out_reset; - } - ret = btrfs_update_root(trans, fs_info->tree_root, &root->root_key, &root->root_item); if (ret < 0) {
Currently when a read-only snapshot is received and subsequently its ro property is set to false i.e. switched to rw-mode the received_uuid of that subvol remains intact. However, once the received volume is switched to RW mode we cannot guaranteee that it contains the same data, so it makes sense to remove the received uuid. The presence of the received_uuid can also cause problems when the volume is being send. Signed-off-by: Nikolay Borisov <nborisov@suse.com> --- v3: * Rework the patch considering latest feedback from David Sterba i.e. explicitly use btrfs_end_transaction fs/btrfs/ioctl.c | 36 +++++++++++++++++++++++++++++------- 1 file changed, 29 insertions(+), 7 deletions(-)