From mboxrd@z Thu Jan 1 00:00:00 1970 From: Lubos Uhliarik Subject: Re: [RFC][PATCH 1/1] ext4: Undelete Feature for Ext4 Date: Fri, 18 Apr 2014 17:48:57 +0200 Message-ID: <1397836137.21164.4.camel@zerobox.home> References: <1395155384.15587.10.camel@zerobox.home> <2C2D46FC-AB12-41F2-BF92-427FD5965C26@dilger.ca> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: linux-ext4@vger.kernel.org, vojnar@fit.vutbr.cz, lczerner@redhat.com To: Andreas Dilger Return-path: Received: from smtp1.seznam.cz ([77.75.72.43]:58322 "EHLO smtp1.seznam.cz" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751242AbaDRPtM (ORCPT ); Fri, 18 Apr 2014 11:49:12 -0400 In-Reply-To: <2C2D46FC-AB12-41F2-BF92-427FD5965C26@dilger.ca> Sender: linux-ext4-owner@vger.kernel.org List-ID: Hi Andreas, I will try to create matching patch for e2fsprogs debugfs later, after = I will finish my patch for ext4.=20 You wrote, that you think inodes have their blocks zeroed out by kernel during truncation, but in fact, it's not happening (I tested it and there are only changes in extent header and ext4_extents structs... Extent_indexes are untached by delete process. Lubos Andreas Dilger p=C3=AD=C5=A1e v =C3=9At 18. 03. 2014 v 12:23 -0600: > On Mar 18, 2014, at 9:09 AM, Lubos Uhliarik wrot= e: > > This patch should make undelete process for deleted files under ext= 4 fs easier. >=20 > Hi Lubos, > thanks for the interesting patch. This does seem like an improvement= =2E > It would be good to add more description to your commit comment, sinc= e > it should explain what is actually being done in your patch (e.g. sto= re > the extent count and depth in the ext4_extent_header eh_generation fi= eld, > ...). >=20 > It would also be good to include a matching patch for e2fsprogs debug= fs > in the "undelete" command. I think one obstacle is the fact that ino= des > have their blocks zeroed out by the kernel during truncation. It wou= ld > be necessary to change the kernel slightly to avoid writing out i_blo= cks > with zero if the inode is being deleted. Otherwise, this change is n= ot > as useful as it could be. >=20 > Cheers, Andreas >=20 > > Signed-off-by: Lubos Uhliarik > > --- > > fs/ext4/ext4_extents.h | 14 +++++++++++ > > fs/ext4/extents.c | 67 +++++++++++++++++++++++++++++++++++++++= ++++++----- > > 2 files changed, 75 insertions(+), 6 deletions(-) > >=20 > > diff --git a/fs/ext4/ext4_extents.h b/fs/ext4/ext4_extents.h > > index 5074fe2..22cf2cd 100644 > > --- a/fs/ext4/ext4_extents.h > > +++ b/fs/ext4/ext4_extents.h > > @@ -251,6 +251,20 @@ static inline void ext4_ext_store_pblock(struc= t ext4_extent *ex, > > 0xffff); > > } > >=20 > > +static inline void ext4_ext_store_entries(struct ext4_extent_heade= r *eh, > > + __u16 eh_entries){ > > + eh->eh_generation &=3D cpu_to_le32((unsigned long) 0x0000ffff); > > + eh->eh_generation |=3D cpu_to_le32(((unsigned long) (eh_entries) = << 16) > > + & 0xffff0000); > > +} > > + > > +static inline void ext4_ext_store_depth(struct ext4_extent_header = *eh, > > + __u16 eh_depth){ > > + eh->eh_generation &=3D cpu_to_le32((unsigned long) 0xffff0000); > > + eh->eh_generation |=3D cpu_to_le32(((unsigned long) (eh_depth)) > > + & 0x0000ffff); > > +} > > + > > /* > > * ext4_idx_store_pblock: > > * stores a large physical block number into an index struct, > > diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c > > index 74bc2d5..f0f3615 100644 > > --- a/fs/ext4/extents.c > > +++ b/fs/ext4/extents.c > > @@ -2565,6 +2565,7 @@ ext4_ext_rm_leaf(handle_t *handle, struct ino= de *inode, > > ext4_lblk_t ex_ee_block; > > unsigned short ex_ee_len; > > unsigned uninitialized =3D 0; > > + unsigned short ex_ee_entries; > > struct ext4_extent *ex; > > ext4_fsblk_t pblk; > >=20 > > @@ -2584,6 +2585,7 @@ ext4_ext_rm_leaf(handle_t *handle, struct ino= de *inode, > >=20 > > ex_ee_block =3D le32_to_cpu(ex->ee_block); > > ex_ee_len =3D ext4_ext_get_actual_len(ex); > > + ex_ee_entries =3D le16_to_cpu(eh->eh_entries); > >=20 > > trace_ext4_ext_rm_leaf(inode, start, ex, *partial_cluster); > >=20 > > @@ -2662,11 +2664,9 @@ ext4_ext_rm_leaf(handle_t *handle, struct in= ode *inode, > > if (err) > > goto out; > >=20 > > - if (num =3D=3D 0) > > - /* this extent is removed; mark slot entirely unused */ > > - ext4_ext_store_pblock(ex, 0); > > + if (num !=3D 0) > > + ex->ee_len =3D cpu_to_le16(num); > >=20 > > - ex->ee_len =3D cpu_to_le16(num); > > /* > > * Do not mark uninitialized if all the blocks in the > > * extent have been removed. > > @@ -2726,8 +2726,19 @@ ext4_ext_rm_leaf(handle_t *handle, struct in= ode *inode, > >=20 > > /* if this leaf is free, then we should > > * remove it from index block above */ > > - if (err =3D=3D 0 && eh->eh_entries =3D=3D 0 && path[depth].p_bh != =3D NULL) > > + if (err =3D=3D 0 && eh->eh_entries =3D=3D 0 && path[depth].p_bh != =3D NULL) { > > + err =3D ext4_ext_get_access(handle, inode, path + depth); > > + if (err) > > + goto out; > > + > > + ext4_ext_store_entries(path[depth].p_hdr, ex_ee_entries); > > + > > + err =3D ext4_ext_dirty(handle, inode, path + depth); > > + if (err) > > + goto out; > > + > > err =3D ext4_ext_rm_idx(handle, inode, path, depth); > > + } > >=20 > > out: > > return err; > > @@ -2760,6 +2771,7 @@ int ext4_ext_remove_space(struct inode *inode= , ext4_lblk_t start, > > struct super_block *sb =3D inode->i_sb; > > int depth =3D ext_depth(inode); > > struct ext4_ext_path *path =3D NULL; > > + unsigned short *entries_path =3D NULL; > > long long partial_cluster =3D 0; > > handle_t *handle; > > int i =3D 0, err =3D 0; > > @@ -2864,6 +2876,18 @@ again: > > } > > err =3D 0; > >=20 > > + if (!inode->i_nlink) { > > + entries_path =3D kzalloc(sizeof(short) * (depth + 1), > > + GFP_NOFS); > > + if (entries_path =3D=3D NULL) { > > + kfree(path); > > + ext4_journal_stop(handle); > > + return -ENOMEM; > > + } > > + > > + entries_path[0] =3D le16_to_cpu(ext_inode_hdr(inode)->eh_entries= ); > > + } > > + > > while (i >=3D 0 && err =3D=3D 0) { > > if (i =3D=3D depth) { > > /* this is leaf block */ > > @@ -2890,6 +2914,10 @@ again: > > ext_debug("init index ptr: hdr 0x%p, num %d\n", > > path[i].p_hdr, > > le16_to_cpu(path[i].p_hdr->eh_entries)); > > + if (entries_path !=3D NULL) { > > + entries_path[i] =3D le16_to_cpu( > > + path[i].p_hdr->eh_entries); > > + } > > } else { > > /* we were already here, see at next index */ > > path[i].p_idx--; > > @@ -2928,10 +2956,24 @@ again: > > } else { > > /* we finished processing this index, go up */ > > if (path[i].p_hdr->eh_entries =3D=3D 0 && i > 0) { > > + if (entries_path !=3D NULL) { > > + err =3D ext4_ext_get_access(handle, > > + inode, path); > > + if (err =3D=3D 0) { > > + ext4_ext_store_entries( > > + path[i].p_hdr, > > + (entries_path[i])); > > + err =3D ext4_ext_dirty(handle, > > + inode, path); > > + } > > + } > > + > > /* index is empty, remove it; > > * handle must be already prepared by the > > * truncatei_leaf() */ > > - err =3D ext4_ext_rm_idx(handle, inode, path, i); > > + if (err =3D=3D 0) > > + err =3D ext4_ext_rm_idx(handle, inode, > > + path, i); > > } > > /* root level has p_bh =3D=3D NULL, brelse() eats this */ > > brelse(path[i].p_bh); > > @@ -2964,6 +3006,17 @@ again: > > */ > > err =3D ext4_ext_get_access(handle, inode, path); > > if (err =3D=3D 0) { > > + /* Store depth and entries to eh_generation to make > > + * recovering deleted files easier. */ > > + if (entries_path !=3D NULL) { > > + ext4_ext_store_entries(ext_inode_hdr(inode), > > + (entries_path[0])); > > + > > + ext4_ext_store_depth(ext_inode_hdr(inode), > > + le16_to_cpu( > > + ext_inode_hdr(inode)->eh_depth)); > > + } > > + > > ext_inode_hdr(inode)->eh_depth =3D 0; > > ext_inode_hdr(inode)->eh_max =3D > > cpu_to_le16(ext4_ext_space_root(inode, 0)); > > @@ -2973,6 +3026,8 @@ again: > > out: > > ext4_ext_drop_refs(path); > > kfree(path); > > + kfree(entries_path); > > + > > if (err =3D=3D -EAGAIN) { > > path =3D NULL; > > goto again; > > --=20 > > 1.8.3.2 > >=20 > >=20 > >=20 > >=20 > >=20 > >=20 > >=20 > > -- > > To unsubscribe from this list: send the line "unsubscribe linux-ext= 4" in > > the body of a message to majordomo@vger.kernel.org > > More majordomo info at http://vger.kernel.org/majordomo-info.html >=20 >=20 > Cheers, Andreas >=20 >=20 >=20 >=20 >=20 -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html