On Thu, Sep 15, 2011 at 11:12:22AM -0400, Theodore Ts'o wrote:
>
> Hi Darrick, Amir,
>
> Could you please take a look at the changes here and make sure they look
> sane to you? I just want to make sure we're all on the same page as
> far as on-disk field assignments are concerned.
>
> Also, one thought for Darrick. I note that with the assignment of
> l_i_checksum, we've exhausted the very last field in the base 128-byte
> inode. Given that the checksum is protecting a 128 byte or 256 byte
> inode, I wonder if we need to use a full 32 bit checksum. Maybe we
> should just use 16 bits of a crc32c?
>
> I did some web searching on the issue, and the recommendation I've come
> across is the following:
>
> "Generally speaking, an n-bit CRC's error detection properties
> degrade after 2**(n-1)-1 data bits."
>
> So a CRC-16 will be good up to just under 4k. A 16-bit truncated crc32c
> will presumably not be as good as a crc16, but seems to me that it's
> probably fine for a 128-256 byte inode. Especially since the main thing
> we're generally worried about is detecting a block getting written to
> the wrong location, overwriting an existing inode table block. So if we
> were really paranoid we could verify the checksums for all of the inodes
> in a particular inode table block when we read in the inode table block
> in question.
On the other hand, you can set inode_size = block_size, which means that with a
4k inode + 32-bit inode number + 16-byte UUID you actually could run afoul of
that degradation. But that seems like an extreme argument for an infrequent
case.
Actually, I've started wondering if we could split the 4 bytes of the crc32c
among the first few inodes of the block, and compute the checksums at block
size granularity. Though that would make inode updates particularly more
expensive... but if I'm going to shift the write-time checksum to a journal
callback then it's not going to matter (for the journal-using users, anyway).
Though with that scheme, you'd probably lose more inodes for any given
integrity error. It also means that the checksum size in each inode becomes
variable (32 bits if inode=blocksize, 16 if inode=blocksize/2, and 8
otherwise), which is a somewhat confusing schema.
<shrug> Do you anticipate a need to add more fields to 128-byte inode
filesystems? I think most of those would be former ext2/3 filesystems,
floppies, and "small" filesystems, correct?
Or does this second scheme sound more attractive?
--D
>
> - Ted
>
>
> commit ceade753f14f2697d329f71b5277b49fd46fcb55
> Author: Theodore Ts'o <[email protected]>
> Date: Thu Sep 15 10:38:55 2011 -0400
>
> libext2fs: add metadata checksum and snapshot feature flags
>
> Reserve EXT4_FEATURE_RO_COMPAT_METADATA_CSUM and
> EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP. Also reserve fields in the
> superblock and the inode for the checksums. In the block group
> descriptor, reserve the exclude bitmap field for the snapshot feature,
> and checksums for the inode and block allocation bitmaps.
>
> With this commit, the metadata checksum and exclude bitmap features
> should have reserved all of the fields they need in ext4's on-disk
> format.
>
> This commit also fixes an a missing byte swap for s_overhead_blocks.
>
> Signed-off-by: "Theodore Ts'o" <[email protected]>
> Cc: Darrick J. Wong <[email protected]>
> Cc: Amir Goldstein <[email protected]>
>
> diff --git a/debugfs/set_fields.c b/debugfs/set_fields.c
> index ac6bc25..cba9c12 100644
> --- a/debugfs/set_fields.c
> +++ b/debugfs/set_fields.c
> @@ -144,6 +144,7 @@ static struct field_set_info super_fields[] = {
> { "usr_quota_inum", &set_sb.s_usr_quota_inum, 4, parse_uint },
> { "grp_quota_inum", &set_sb.s_grp_quota_inum, 4, parse_uint },
> { "overhead_blocks", &set_sb.s_overhead_blocks, 4, parse_uint },
> + { "checksum", &set_sb.s_checksum, 2, parse_uint },
> { 0, 0, 0, 0 }
> };
>
> @@ -179,6 +180,7 @@ static struct field_set_info inode_fields[] = {
> { "fsize", &set_inode.osd2.hurd2.h_i_fsize, 1, parse_uint },
> { "uid_high", &set_inode.osd2.linux2.l_i_uid_high, 2, parse_uint },
> { "gid_high", &set_inode.osd2.linux2.l_i_gid_high, 2, parse_uint },
> + { "checksum", &set_inode.osd2.linux2.l_i_checksum, 4, parse_uint },
> { "author", &set_inode.osd2.hurd2.h_i_author, 4, parse_uint },
> { "bmap", NULL, 4, parse_bmap, FLAG_ARRAY },
> { 0, 0, 0, 0 }
> @@ -192,7 +194,6 @@ static struct field_set_info ext2_bg_fields[] = {
> { "free_inodes_count", &set_gd.bg_free_inodes_count, 2, parse_uint },
> { "used_dirs_count", &set_gd.bg_used_dirs_count, 2, parse_uint },
> { "flags", &set_gd.bg_flags, 2, parse_uint },
> - { "reserved", &set_gd.bg_reserved, 2, parse_uint, FLAG_ARRAY, 2 },
> { "itable_unused", &set_gd.bg_itable_unused, 2, parse_uint },
> { "checksum", &set_gd.bg_checksum, 2, parse_gd_csum },
> { 0, 0, 0, 0 }
> diff --git a/lib/e2p/feature.c b/lib/e2p/feature.c
> index 16fba53..965fc16 100644
> --- a/lib/e2p/feature.c
> +++ b/lib/e2p/feature.c
> @@ -40,6 +40,8 @@ static struct feature feature_list[] = {
> "resize_inode" },
> { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_LAZY_BG,
> "lazy_bg" },
> + { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP,
> + "snapshot" },
>
> { E2P_FEATURE_RO_INCOMPAT, EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER,
> "sparse_super" },
> @@ -59,6 +61,8 @@ static struct feature feature_list[] = {
> "quota" },
> { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_BIGALLOC,
> "bigalloc"},
> + { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_METADATA_CSUM,
> + "metadata_csum"},
>
> { E2P_FEATURE_INCOMPAT, EXT2_FEATURE_INCOMPAT_COMPRESSION,
> "compression" },
> diff --git a/lib/e2p/ls.c b/lib/e2p/ls.c
> index 0f36f40..aaacdaa 100644
> --- a/lib/e2p/ls.c
> +++ b/lib/e2p/ls.c
> @@ -413,6 +413,10 @@ void list_super2(struct ext2_super_block * sb, FILE *f)
> if (sb->s_grp_quota_inum)
> fprintf(f, "Group quota inode: %u\n",
> sb->s_grp_quota_inum);
> +
> + if (sb->s_feature_ro_compat & EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)
> + fprintf(f, "Checksum: 0x%08x\n",
> + sb->s_checksum);
> }
>
> void list_super (struct ext2_super_block * s)
> diff --git a/lib/ext2fs/ext2_fs.h b/lib/ext2fs/ext2_fs.h
> index 4fec5db..1b02054 100644
> --- a/lib/ext2fs/ext2_fs.h
> +++ b/lib/ext2fs/ext2_fs.h
> @@ -142,7 +142,9 @@ struct ext2_group_desc
> __u16 bg_free_inodes_count; /* Free inodes count */
> __u16 bg_used_dirs_count; /* Directories count */
> __u16 bg_flags;
> - __u32 bg_reserved[2];
> + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
> + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> __u16 bg_itable_unused; /* Unused inodes count */
> __u16 bg_checksum; /* crc16(s_uuid+grouo_num+group_desc)*/
> };
> @@ -159,7 +161,9 @@ struct ext4_group_desc
> __u16 bg_free_inodes_count; /* Free inodes count */
> __u16 bg_used_dirs_count; /* Directories count */
> __u16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
> - __u32 bg_reserved[2]; /* Likely block/inode bitmap checksum */
> + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
> + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> __u16 bg_itable_unused; /* Unused inodes count */
> __u16 bg_checksum; /* crc16(sb_uuid+group+desc) */
> __u32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
> @@ -169,7 +173,10 @@ struct ext4_group_desc
> __u16 bg_free_inodes_count_hi;/* Free inodes count MSB */
> __u16 bg_used_dirs_count_hi; /* Directories count MSB */
> __u16 bg_itable_unused_hi; /* Unused inodes count MSB */
> - __u32 bg_reserved2[3];
> + __u32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
> + __u16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
> + __u16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
> + __u32 bg_reserved;
> };
>
> #define EXT2_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not initialized */
> @@ -363,7 +370,7 @@ struct ext2_inode {
> __u16 l_i_file_acl_high;
> __u16 l_i_uid_high; /* these 2 fields */
> __u16 l_i_gid_high; /* were reserved2[0] */
> - __u32 l_i_reserved2;
> + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
> } linux2;
> struct {
> __u8 h_i_frag; /* Fragment number */
> @@ -410,7 +417,7 @@ struct ext2_inode_large {
> __u16 l_i_file_acl_high;
> __u16 l_i_uid_high; /* these 2 fields */
> __u16 l_i_gid_high; /* were reserved2[0] */
> - __u32 l_i_reserved2;
> + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
> } linux2;
> struct {
> __u8 h_i_frag; /* Fragment number */
> @@ -441,7 +448,7 @@ struct ext2_inode_large {
> #define i_gid_low i_gid
> #define i_uid_high osd2.linux2.l_i_uid_high
> #define i_gid_high osd2.linux2.l_i_gid_high
> -#define i_reserved2 osd2.linux2.l_i_reserved2
> +#define i_checksum osd2.linux2.l_i_checksum
> #else
> #if defined(__GNU__)
>
> @@ -623,7 +630,8 @@ struct ext2_super_block {
> __u32 s_usr_quota_inum; /* inode number of user quota file */
> __u32 s_grp_quota_inum; /* inode number of group quota file */
> __u32 s_overhead_blocks; /* overhead blocks/clusters in fs */
> - __u32 s_reserved[109]; /* Padding to the end of the block */
> + __u32 s_checksum; /* crc32c(superblock) */
> + __u32 s_reserved[108]; /* Padding to the end of the block */
> };
>
> #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
> @@ -671,7 +679,9 @@ struct ext2_super_block {
> #define EXT2_FEATURE_COMPAT_RESIZE_INODE 0x0010
> #define EXT2_FEATURE_COMPAT_DIR_INDEX 0x0020
> #define EXT2_FEATURE_COMPAT_LAZY_BG 0x0040
> -#define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080
> +/* #define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080 not used, legacy */
> +#define EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP 0x0100
> +
>
> #define EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
> #define EXT2_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
> @@ -683,6 +693,7 @@ struct ext2_super_block {
> #define EXT4_FEATURE_RO_COMPAT_HAS_SNAPSHOT 0x0080
> #define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
> #define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
> +#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
>
> #define EXT2_FEATURE_INCOMPAT_COMPRESSION 0x0001
> #define EXT2_FEATURE_INCOMPAT_FILETYPE 0x0002
> diff --git a/lib/ext2fs/swapfs.c b/lib/ext2fs/swapfs.c
> index 87b1a2e..d1c4a56 100644
> --- a/lib/ext2fs/swapfs.c
> +++ b/lib/ext2fs/swapfs.c
> @@ -78,6 +78,8 @@ void ext2fs_swap_super(struct ext2_super_block * sb)
> sb->s_snapshot_list = ext2fs_swab32(sb->s_snapshot_list);
> sb->s_usr_quota_inum = ext2fs_swab32(sb->s_usr_quota_inum);
> sb->s_grp_quota_inum = ext2fs_swab32(sb->s_grp_quota_inum);
> + sb->s_overhead_blocks = ext2fs_swab32(sb->s_overhead_blocks);
> + sb->s_checksum = ext2fs_swab32(sb->s_checksum);
>
> for (i=0; i < 4; i++)
> sb->s_hash_seed[i] = ext2fs_swab32(sb->s_hash_seed[i]);
> @@ -106,6 +108,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
> gdp->bg_free_inodes_count = ext2fs_swab16(gdp->bg_free_inodes_count);
> gdp->bg_used_dirs_count = ext2fs_swab16(gdp->bg_used_dirs_count);
> gdp->bg_flags = ext2fs_swab16(gdp->bg_flags);
> + gdp->bg_exclude_bitmap_lo = ext2fs_swab32(gdp->bg_exclude_bitmap_lo);
> + gdp->bg_block_bitmap_csum_lo =
> + ext2fs_swab16(gdp->bg_block_bitmap_csum_lo);
> + gdp->bg_inode_bitmap_csum_lo =
> + ext2fs_swab16(gdp->bg_inode_bitmap_csum_lo);
> gdp->bg_itable_unused = ext2fs_swab16(gdp->bg_itable_unused);
> gdp->bg_checksum = ext2fs_swab16(gdp->bg_checksum);
> /* If we're 32-bit, we're done */
> @@ -125,6 +132,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
> gdp4->bg_used_dirs_count_hi =
> ext2fs_swab16(gdp4->bg_used_dirs_count_hi);
> gdp4->bg_itable_unused_hi = ext2fs_swab16(gdp4->bg_itable_unused_hi);
> + gdp->bg_exclude_bitmap_hi = ext2fs_swab16(gdp->bg_exclude_bitmap_hi);
> + gdp->bg_block_bitmap_csum_hi =
> + ext2fs_swab16(gdp->bg_block_bitmap_csum_hi);
> + gdp->bg_inode_bitmap_csum_hi =
> + ext2fs_swab16(gdp->bg_inode_bitmap_csum_hi);
> }
>
> void ext2fs_swap_group_desc(struct ext2_group_desc *gdp)
> @@ -244,8 +256,8 @@ void ext2fs_swap_inode_full(ext2_filsys fs, struct ext2_inode_large *t,
> ext2fs_swab16 (f->osd2.linux2.l_i_uid_high);
> t->osd2.linux2.l_i_gid_high =
> ext2fs_swab16 (f->osd2.linux2.l_i_gid_high);
> - t->osd2.linux2.l_i_reserved2 =
> - ext2fs_swab32(f->osd2.linux2.l_i_reserved2);
> + t->osd2.linux2.l_i_checksum =
> + ext2fs_swab32(f->osd2.linux2.checksum);
> break;
> case EXT2_OS_HURD:
> t->osd1.hurd1.h_i_translator =
> diff --git a/lib/ext2fs/tst_inode_size.c b/lib/ext2fs/tst_inode_size.c
> index 962f1cd..683b79c 100644
> --- a/lib/ext2fs/tst_inode_size.c
> +++ b/lib/ext2fs/tst_inode_size.c
> @@ -61,7 +61,7 @@ void check_structure_fields()
> check_field(osd2.linux2.l_i_file_acl_high);
> check_field(osd2.linux2.l_i_uid_high);
> check_field(osd2.linux2.l_i_gid_high);
> - check_field(osd2.linux2.l_i_reserved2);
> + check_field(osd2.linux2.l_i_checksum);
> printf("Ending offset is %d\n\n", cur_offset);
> #endif
> }
> diff --git a/lib/ext2fs/tst_super_size.c b/lib/ext2fs/tst_super_size.c
> index 1e5a524..75659ae 100644
> --- a/lib/ext2fs/tst_super_size.c
> +++ b/lib/ext2fs/tst_super_size.c
> @@ -126,6 +126,7 @@ void check_superblock_fields()
> check_field(s_usr_quota_inum);
> check_field(s_grp_quota_inum);
> check_field(s_overhead_blocks);
> + check_field(s_checksum);
> check_field(s_reserved);
> printf("Ending offset is %d\n\n", cur_offset);
> #endif
>
On Thu, Sep 15, 2011 at 09:55:12AM -0700, Darrick J. Wong wrote:
> On Thu, Sep 15, 2011 at 11:12:22AM -0400, Theodore Ts'o wrote:
> >
> > Hi Darrick, Amir,
> >
> > Could you please take a look at the changes here and make sure they look
> > sane to you? I just want to make sure we're all on the same page as
> > far as on-disk field assignments are concerned.
> >
> > Also, one thought for Darrick. I note that with the assignment of
> > l_i_checksum, we've exhausted the very last field in the base 128-byte
> > inode. Given that the checksum is protecting a 128 byte or 256 byte
> > inode, I wonder if we need to use a full 32 bit checksum. Maybe we
> > should just use 16 bits of a crc32c?
> >
> > I did some web searching on the issue, and the recommendation I've come
> > across is the following:
> >
> > "Generally speaking, an n-bit CRC's error detection properties
> > degrade after 2**(n-1)-1 data bits."
> >
> > So a CRC-16 will be good up to just under 4k. A 16-bit truncated crc32c
> > will presumably not be as good as a crc16, but seems to me that it's
> > probably fine for a 128-256 byte inode. Especially since the main thing
> > we're generally worried about is detecting a block getting written to
> > the wrong location, overwriting an existing inode table block. So if we
> > were really paranoid we could verify the checksums for all of the inodes
> > in a particular inode table block when we read in the inode table block
> > in question.
>
> On the other hand, you can set inode_size = block_size, which means that with a
> 4k inode + 32-bit inode number + 16-byte UUID you actually could run afoul of
> that degradation. But that seems like an extreme argument for an infrequent
> case.
>
> Actually, I've started wondering if we could split the 4 bytes of the crc32c
> among the first few inodes of the block, and compute the checksums at block
> size granularity. Though that would make inode updates particularly more
> expensive... but if I'm going to shift the write-time checksum to a journal
> callback then it's not going to matter (for the journal-using users, anyway).
>
> Though with that scheme, you'd probably lose more inodes for any given
> integrity error. It also means that the checksum size in each inode becomes
> variable (32 bits if inode=blocksize, 16 if inode=blocksize/2, and 8
> otherwise), which is a somewhat confusing schema.
>
> <shrug> Do you anticipate a need to add more fields to 128-byte inode
> filesystems? I think most of those would be former ext2/3 filesystems,
> floppies, and "small" filesystems, correct?
>
> Or does this second scheme sound more attractive?
I forgot to say, "as opposed to storing the lower 16 bits below 128 bytes and
the upper 16 bits somewhere above it."
--D
>
> --D
> >
> > - Ted
> >
> >
> > commit ceade753f14f2697d329f71b5277b49fd46fcb55
> > Author: Theodore Ts'o <[email protected]>
> > Date: Thu Sep 15 10:38:55 2011 -0400
> >
> > libext2fs: add metadata checksum and snapshot feature flags
> >
> > Reserve EXT4_FEATURE_RO_COMPAT_METADATA_CSUM and
> > EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP. Also reserve fields in the
> > superblock and the inode for the checksums. In the block group
> > descriptor, reserve the exclude bitmap field for the snapshot feature,
> > and checksums for the inode and block allocation bitmaps.
> >
> > With this commit, the metadata checksum and exclude bitmap features
> > should have reserved all of the fields they need in ext4's on-disk
> > format.
> >
> > This commit also fixes an a missing byte swap for s_overhead_blocks.
> >
> > Signed-off-by: "Theodore Ts'o" <[email protected]>
> > Cc: Darrick J. Wong <[email protected]>
> > Cc: Amir Goldstein <[email protected]>
> >
> > diff --git a/debugfs/set_fields.c b/debugfs/set_fields.c
> > index ac6bc25..cba9c12 100644
> > --- a/debugfs/set_fields.c
> > +++ b/debugfs/set_fields.c
> > @@ -144,6 +144,7 @@ static struct field_set_info super_fields[] = {
> > { "usr_quota_inum", &set_sb.s_usr_quota_inum, 4, parse_uint },
> > { "grp_quota_inum", &set_sb.s_grp_quota_inum, 4, parse_uint },
> > { "overhead_blocks", &set_sb.s_overhead_blocks, 4, parse_uint },
> > + { "checksum", &set_sb.s_checksum, 2, parse_uint },
> > { 0, 0, 0, 0 }
> > };
> >
> > @@ -179,6 +180,7 @@ static struct field_set_info inode_fields[] = {
> > { "fsize", &set_inode.osd2.hurd2.h_i_fsize, 1, parse_uint },
> > { "uid_high", &set_inode.osd2.linux2.l_i_uid_high, 2, parse_uint },
> > { "gid_high", &set_inode.osd2.linux2.l_i_gid_high, 2, parse_uint },
> > + { "checksum", &set_inode.osd2.linux2.l_i_checksum, 4, parse_uint },
> > { "author", &set_inode.osd2.hurd2.h_i_author, 4, parse_uint },
> > { "bmap", NULL, 4, parse_bmap, FLAG_ARRAY },
> > { 0, 0, 0, 0 }
> > @@ -192,7 +194,6 @@ static struct field_set_info ext2_bg_fields[] = {
> > { "free_inodes_count", &set_gd.bg_free_inodes_count, 2, parse_uint },
> > { "used_dirs_count", &set_gd.bg_used_dirs_count, 2, parse_uint },
> > { "flags", &set_gd.bg_flags, 2, parse_uint },
> > - { "reserved", &set_gd.bg_reserved, 2, parse_uint, FLAG_ARRAY, 2 },
> > { "itable_unused", &set_gd.bg_itable_unused, 2, parse_uint },
> > { "checksum", &set_gd.bg_checksum, 2, parse_gd_csum },
> > { 0, 0, 0, 0 }
> > diff --git a/lib/e2p/feature.c b/lib/e2p/feature.c
> > index 16fba53..965fc16 100644
> > --- a/lib/e2p/feature.c
> > +++ b/lib/e2p/feature.c
> > @@ -40,6 +40,8 @@ static struct feature feature_list[] = {
> > "resize_inode" },
> > { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_LAZY_BG,
> > "lazy_bg" },
> > + { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP,
> > + "snapshot" },
> >
> > { E2P_FEATURE_RO_INCOMPAT, EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER,
> > "sparse_super" },
> > @@ -59,6 +61,8 @@ static struct feature feature_list[] = {
> > "quota" },
> > { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_BIGALLOC,
> > "bigalloc"},
> > + { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_METADATA_CSUM,
> > + "metadata_csum"},
> >
> > { E2P_FEATURE_INCOMPAT, EXT2_FEATURE_INCOMPAT_COMPRESSION,
> > "compression" },
> > diff --git a/lib/e2p/ls.c b/lib/e2p/ls.c
> > index 0f36f40..aaacdaa 100644
> > --- a/lib/e2p/ls.c
> > +++ b/lib/e2p/ls.c
> > @@ -413,6 +413,10 @@ void list_super2(struct ext2_super_block * sb, FILE *f)
> > if (sb->s_grp_quota_inum)
> > fprintf(f, "Group quota inode: %u\n",
> > sb->s_grp_quota_inum);
> > +
> > + if (sb->s_feature_ro_compat & EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)
> > + fprintf(f, "Checksum: 0x%08x\n",
> > + sb->s_checksum);
> > }
> >
> > void list_super (struct ext2_super_block * s)
> > diff --git a/lib/ext2fs/ext2_fs.h b/lib/ext2fs/ext2_fs.h
> > index 4fec5db..1b02054 100644
> > --- a/lib/ext2fs/ext2_fs.h
> > +++ b/lib/ext2fs/ext2_fs.h
> > @@ -142,7 +142,9 @@ struct ext2_group_desc
> > __u16 bg_free_inodes_count; /* Free inodes count */
> > __u16 bg_used_dirs_count; /* Directories count */
> > __u16 bg_flags;
> > - __u32 bg_reserved[2];
> > + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
> > + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> > + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> > __u16 bg_itable_unused; /* Unused inodes count */
> > __u16 bg_checksum; /* crc16(s_uuid+grouo_num+group_desc)*/
> > };
> > @@ -159,7 +161,9 @@ struct ext4_group_desc
> > __u16 bg_free_inodes_count; /* Free inodes count */
> > __u16 bg_used_dirs_count; /* Directories count */
> > __u16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
> > - __u32 bg_reserved[2]; /* Likely block/inode bitmap checksum */
> > + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
> > + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> > + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> > __u16 bg_itable_unused; /* Unused inodes count */
> > __u16 bg_checksum; /* crc16(sb_uuid+group+desc) */
> > __u32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
> > @@ -169,7 +173,10 @@ struct ext4_group_desc
> > __u16 bg_free_inodes_count_hi;/* Free inodes count MSB */
> > __u16 bg_used_dirs_count_hi; /* Directories count MSB */
> > __u16 bg_itable_unused_hi; /* Unused inodes count MSB */
> > - __u32 bg_reserved2[3];
> > + __u32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
> > + __u16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
> > + __u16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
> > + __u32 bg_reserved;
> > };
> >
> > #define EXT2_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not initialized */
> > @@ -363,7 +370,7 @@ struct ext2_inode {
> > __u16 l_i_file_acl_high;
> > __u16 l_i_uid_high; /* these 2 fields */
> > __u16 l_i_gid_high; /* were reserved2[0] */
> > - __u32 l_i_reserved2;
> > + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
> > } linux2;
> > struct {
> > __u8 h_i_frag; /* Fragment number */
> > @@ -410,7 +417,7 @@ struct ext2_inode_large {
> > __u16 l_i_file_acl_high;
> > __u16 l_i_uid_high; /* these 2 fields */
> > __u16 l_i_gid_high; /* were reserved2[0] */
> > - __u32 l_i_reserved2;
> > + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
> > } linux2;
> > struct {
> > __u8 h_i_frag; /* Fragment number */
> > @@ -441,7 +448,7 @@ struct ext2_inode_large {
> > #define i_gid_low i_gid
> > #define i_uid_high osd2.linux2.l_i_uid_high
> > #define i_gid_high osd2.linux2.l_i_gid_high
> > -#define i_reserved2 osd2.linux2.l_i_reserved2
> > +#define i_checksum osd2.linux2.l_i_checksum
> > #else
> > #if defined(__GNU__)
> >
> > @@ -623,7 +630,8 @@ struct ext2_super_block {
> > __u32 s_usr_quota_inum; /* inode number of user quota file */
> > __u32 s_grp_quota_inum; /* inode number of group quota file */
> > __u32 s_overhead_blocks; /* overhead blocks/clusters in fs */
> > - __u32 s_reserved[109]; /* Padding to the end of the block */
> > + __u32 s_checksum; /* crc32c(superblock) */
> > + __u32 s_reserved[108]; /* Padding to the end of the block */
> > };
> >
> > #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
> > @@ -671,7 +679,9 @@ struct ext2_super_block {
> > #define EXT2_FEATURE_COMPAT_RESIZE_INODE 0x0010
> > #define EXT2_FEATURE_COMPAT_DIR_INDEX 0x0020
> > #define EXT2_FEATURE_COMPAT_LAZY_BG 0x0040
> > -#define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080
> > +/* #define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080 not used, legacy */
> > +#define EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP 0x0100
> > +
> >
> > #define EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
> > #define EXT2_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
> > @@ -683,6 +693,7 @@ struct ext2_super_block {
> > #define EXT4_FEATURE_RO_COMPAT_HAS_SNAPSHOT 0x0080
> > #define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
> > #define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
> > +#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
> >
> > #define EXT2_FEATURE_INCOMPAT_COMPRESSION 0x0001
> > #define EXT2_FEATURE_INCOMPAT_FILETYPE 0x0002
> > diff --git a/lib/ext2fs/swapfs.c b/lib/ext2fs/swapfs.c
> > index 87b1a2e..d1c4a56 100644
> > --- a/lib/ext2fs/swapfs.c
> > +++ b/lib/ext2fs/swapfs.c
> > @@ -78,6 +78,8 @@ void ext2fs_swap_super(struct ext2_super_block * sb)
> > sb->s_snapshot_list = ext2fs_swab32(sb->s_snapshot_list);
> > sb->s_usr_quota_inum = ext2fs_swab32(sb->s_usr_quota_inum);
> > sb->s_grp_quota_inum = ext2fs_swab32(sb->s_grp_quota_inum);
> > + sb->s_overhead_blocks = ext2fs_swab32(sb->s_overhead_blocks);
> > + sb->s_checksum = ext2fs_swab32(sb->s_checksum);
> >
> > for (i=0; i < 4; i++)
> > sb->s_hash_seed[i] = ext2fs_swab32(sb->s_hash_seed[i]);
> > @@ -106,6 +108,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
> > gdp->bg_free_inodes_count = ext2fs_swab16(gdp->bg_free_inodes_count);
> > gdp->bg_used_dirs_count = ext2fs_swab16(gdp->bg_used_dirs_count);
> > gdp->bg_flags = ext2fs_swab16(gdp->bg_flags);
> > + gdp->bg_exclude_bitmap_lo = ext2fs_swab32(gdp->bg_exclude_bitmap_lo);
> > + gdp->bg_block_bitmap_csum_lo =
> > + ext2fs_swab16(gdp->bg_block_bitmap_csum_lo);
> > + gdp->bg_inode_bitmap_csum_lo =
> > + ext2fs_swab16(gdp->bg_inode_bitmap_csum_lo);
> > gdp->bg_itable_unused = ext2fs_swab16(gdp->bg_itable_unused);
> > gdp->bg_checksum = ext2fs_swab16(gdp->bg_checksum);
> > /* If we're 32-bit, we're done */
> > @@ -125,6 +132,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
> > gdp4->bg_used_dirs_count_hi =
> > ext2fs_swab16(gdp4->bg_used_dirs_count_hi);
> > gdp4->bg_itable_unused_hi = ext2fs_swab16(gdp4->bg_itable_unused_hi);
> > + gdp->bg_exclude_bitmap_hi = ext2fs_swab16(gdp->bg_exclude_bitmap_hi);
> > + gdp->bg_block_bitmap_csum_hi =
> > + ext2fs_swab16(gdp->bg_block_bitmap_csum_hi);
> > + gdp->bg_inode_bitmap_csum_hi =
> > + ext2fs_swab16(gdp->bg_inode_bitmap_csum_hi);
> > }
> >
> > void ext2fs_swap_group_desc(struct ext2_group_desc *gdp)
> > @@ -244,8 +256,8 @@ void ext2fs_swap_inode_full(ext2_filsys fs, struct ext2_inode_large *t,
> > ext2fs_swab16 (f->osd2.linux2.l_i_uid_high);
> > t->osd2.linux2.l_i_gid_high =
> > ext2fs_swab16 (f->osd2.linux2.l_i_gid_high);
> > - t->osd2.linux2.l_i_reserved2 =
> > - ext2fs_swab32(f->osd2.linux2.l_i_reserved2);
> > + t->osd2.linux2.l_i_checksum =
> > + ext2fs_swab32(f->osd2.linux2.checksum);
> > break;
> > case EXT2_OS_HURD:
> > t->osd1.hurd1.h_i_translator =
> > diff --git a/lib/ext2fs/tst_inode_size.c b/lib/ext2fs/tst_inode_size.c
> > index 962f1cd..683b79c 100644
> > --- a/lib/ext2fs/tst_inode_size.c
> > +++ b/lib/ext2fs/tst_inode_size.c
> > @@ -61,7 +61,7 @@ void check_structure_fields()
> > check_field(osd2.linux2.l_i_file_acl_high);
> > check_field(osd2.linux2.l_i_uid_high);
> > check_field(osd2.linux2.l_i_gid_high);
> > - check_field(osd2.linux2.l_i_reserved2);
> > + check_field(osd2.linux2.l_i_checksum);
> > printf("Ending offset is %d\n\n", cur_offset);
> > #endif
> > }
> > diff --git a/lib/ext2fs/tst_super_size.c b/lib/ext2fs/tst_super_size.c
> > index 1e5a524..75659ae 100644
> > --- a/lib/ext2fs/tst_super_size.c
> > +++ b/lib/ext2fs/tst_super_size.c
> > @@ -126,6 +126,7 @@ void check_superblock_fields()
> > check_field(s_usr_quota_inum);
> > check_field(s_grp_quota_inum);
> > check_field(s_overhead_blocks);
> > + check_field(s_checksum);
> > check_field(s_reserved);
> > printf("Ending offset is %d\n\n", cur_offset);
> > #endif
> >
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
On Thu, Sep 15, 2011 at 09:55:12AM -0700, Darrick J. Wong wrote:
> On the other hand, you can set inode_size = block_size, which means
> that with a 4k inode + 32-bit inode number + 16-byte UUID you
> actually could run afoul of that degradation. But that seems like
> an extreme argument for an infrequent case.
Yeah, that's an extremely infrequent case; in fact, I doubt it would
occur much besides testing, and as I've said, the main interest I have
for doing checksums is to detect gross defects, not subtle ones. (One
and two bit errors will be caught by disk drive.)
> <shrug> Do you anticipate a need to add more fields to 128-byte inode
> filesystems? I think most of those would be former ext2/3 filesystems,
> floppies, and "small" filesystems, correct?
Actually, at $WORK we're still using 128 byte inodes. If you don't
need the high resolution timestamps or fast access to extended
attributes, there's no real point to using 256 byte inodes. And 128
byte inodes allow you to pack more inodes per block, which can make
for a noticeable performance difference. It's just since Fedora turns
on SELinux by default, the per-inode labels stored in xattrs suck so
much performance if you don't use 256 byte inods that most people
don't notice the performance degredation going from 128->256 byte
inodes.
> Actually, I've started wondering if we could split the 4 bytes of the crc32c
> among the first few inodes of the block, and compute the checksums at block
> size granularity. Though that would make inode updates particularly more
> expensive... but if I'm going to shift the write-time checksum to a journal
> callback then it's not going to matter (for the journal-using users, anyway).
Yeah, I'd like to keep things cheap for the non-journal case; it's not
just at Google; anyone using ext4 where data reliability is being
handled via replication or reed-solomon encoding at the cluster file
system level (and Hadoopfs does this) is very likely going to be
interested in ext4 w/o a journal.
> Though with that scheme, you'd probably lose more inodes for any given
> integrity error. It also means that the checksum size in each inode becomes
> variable (32 bits if inode=blocksize, 16 if inode=blocksize/2, and 8
> otherwise), which is a somewhat confusing schema.
On a disk drive, in practice the unit of data getting garbled is on a
per-sector basis. It's highly, highly unlikely that that the first
128 bytes will be garbaged, and the second 128 bytes will be OK. I
suppose that could happen if things got corrupted in memory, but
that's what ECC memory is for, right? :-)
- Ted
On 2011-09-15, at 11:19 AM, Darrick J. Wong wrote:
> On Thu, Sep 15, 2011 at 09:55:12AM -0700, Darrick J. Wong wrote:
>> On Thu, Sep 15, 2011 at 11:12:22AM -0400, Theodore Ts'o wrote:
>>>
>>> Hi Darrick, Amir,
>>>
>>> Could you please take a look at the changes here and make sure they look
>>> sane to you? I just want to make sure we're all on the same page as
>>> far as on-disk field assignments are concerned.
>>>
>>> Also, one thought for Darrick. I note that with the assignment of
>>> l_i_checksum, we've exhausted the very last field in the base 128-byte
>>> inode. Given that the checksum is protecting a 128 byte or 256 byte
>>> inode, I wonder if we need to use a full 32 bit checksum. Maybe we
>>> should just use 16 bits of a crc32c?
>>>
>>> I did some web searching on the issue, and the recommendation I've come
>>> across is the following:
>>>
>>> "Generally speaking, an n-bit CRC's error detection properties
>>> degrade after 2**(n-1)-1 data bits."
>>>
>>> So a CRC-16 will be good up to just under 4k. A 16-bit truncated crc32c
>>> will presumably not be as good as a crc16, but seems to me that it's
>>> probably fine for a 128-256 byte inode. Especially since the main thing
>>> we're generally worried about is detecting a block getting written to
>>> the wrong location, overwriting an existing inode table block. So if we
>>> were really paranoid we could verify the checksums for all of the inodes
>>> in a particular inode table block when we read in the inode table block
>>> in question.
>>
>> On the other hand, you can set inode_size = block_size, which means that
>> with a 4k inode + 32-bit inode number + 16-byte UUID you actually could
>> run afoul of that degradation. But that seems like an extreme argument
>> for an infrequent case.
>>
>> Actually, I've started wondering if we could split the 4 bytes of the crc32c
>> among the first few inodes of the block, and compute the checksums at block
>> size granularity. Though that would make inode updates particularly more
>> expensive... but if I'm going to shift the write-time checksum to a journal
>> callback then it's not going to matter (for the journal-using users, anyway).
>>
>> Though with that scheme, you'd probably lose more inodes for any given
>> integrity error. It also means that the checksum size in each inode becomes
>> variable (32 bits if inode=blocksize, 16 if inode=blocksize/2, and 8
>> otherwise), which is a somewhat confusing schema.
>>
>> <shrug> Do you anticipate a need to add more fields to 128-byte inode
>> filesystems? I think most of those would be former ext2/3 filesystems,
>> floppies, and "small" filesystems, correct?
>>
>> Or does this second scheme sound more attractive?
>
> I forgot to say, "as opposed to storing the lower 16 bits below 128 bytes and
> the upper 16 bits somewhere above it."
This is also a possible alternative, though it makes for more fragments that
need to be checksummed. I think as a general rule it makes sense to store
the checksum as the last word in the structure, if possible, so that the
checksum can be computed in a single call. This is already done for 128-byte
inodes and for 32-byte group descriptors, but should also be done for the
s_checksum field in the superblock (i.e. put it after s_reserved instead of
before).
For inodes 256-bytes or larger (which are commonly used for Lustre to store
large xattrs that are needed for every file access), and 64-byte group
descriptors it has to checksum 2 fragments, but at least not more than that
if we precompute for each inode the crc32c(uuid + inum) seed and for each
group the crc32c(uuid + group) seed. The superblock already contains the
UUID, but it may make sense to still precompute the crc32c(uuid) part and
store it in ext4_sb_info for computing the inode seed.
It really would be interesting to measure the crc32c() and crc16() performance
for 512MB in chunks of 4, 32, 128, 256, and 4096 bytes (which is the largest
that we will generally use until we get to data checksums). That would give
us a good idea how fast the checksums _really_ are in our actual usage.
>>> commit ceade753f14f2697d329f71b5277b49fd46fcb55
>>> Author: Theodore Ts'o <[email protected]>
>>> Date: Thu Sep 15 10:38:55 2011 -0400
>>>
>>> libext2fs: add metadata checksum and snapshot feature flags
>>>
>>> Reserve EXT4_FEATURE_RO_COMPAT_METADATA_CSUM and
>>> EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP. Also reserve fields in the
>>> superblock and the inode for the checksums. In the block group
>>> descriptor, reserve the exclude bitmap field for the snapshot feature,
>>> and checksums for the inode and block allocation bitmaps.
>>>
>>> With this commit, the metadata checksum and exclude bitmap features
>>> should have reserved all of the fields they need in ext4's on-disk
>>> format.
>>>
>>> This commit also fixes an a missing byte swap for s_overhead_blocks.
>>>
>>> Signed-off-by: "Theodore Ts'o" <[email protected]>
>>> Cc: Darrick J. Wong <[email protected]>
>>> Cc: Amir Goldstein <[email protected]>
>>>
>>> diff --git a/debugfs/set_fields.c b/debugfs/set_fields.c
>>> index ac6bc25..cba9c12 100644
>>> --- a/debugfs/set_fields.c
>>> +++ b/debugfs/set_fields.c
>>> @@ -144,6 +144,7 @@ static struct field_set_info super_fields[] = {
>>> { "usr_quota_inum", &set_sb.s_usr_quota_inum, 4, parse_uint },
>>> { "grp_quota_inum", &set_sb.s_grp_quota_inum, 4, parse_uint },
>>> { "overhead_blocks", &set_sb.s_overhead_blocks, 4, parse_uint },
>>> + { "checksum", &set_sb.s_checksum, 2, parse_uint },
>>> { 0, 0, 0, 0 }
>>> };
>>>
>>> @@ -179,6 +180,7 @@ static struct field_set_info inode_fields[] = {
>>> { "fsize", &set_inode.osd2.hurd2.h_i_fsize, 1, parse_uint },
>>> { "uid_high", &set_inode.osd2.linux2.l_i_uid_high, 2, parse_uint },
>>> { "gid_high", &set_inode.osd2.linux2.l_i_gid_high, 2, parse_uint },
>>> + { "checksum", &set_inode.osd2.linux2.l_i_checksum, 4, parse_uint },
>>> { "author", &set_inode.osd2.hurd2.h_i_author, 4, parse_uint },
>>> { "bmap", NULL, 4, parse_bmap, FLAG_ARRAY },
>>> { 0, 0, 0, 0 }
>>> @@ -192,7 +194,6 @@ static struct field_set_info ext2_bg_fields[] = {
>>> { "free_inodes_count", &set_gd.bg_free_inodes_count, 2, parse_uint },
>>> { "used_dirs_count", &set_gd.bg_used_dirs_count, 2, parse_uint },
>>> { "flags", &set_gd.bg_flags, 2, parse_uint },
>>> - { "reserved", &set_gd.bg_reserved, 2, parse_uint, FLAG_ARRAY, 2 },
>>> { "itable_unused", &set_gd.bg_itable_unused, 2, parse_uint },
>>> { "checksum", &set_gd.bg_checksum, 2, parse_gd_csum },
>>> { 0, 0, 0, 0 }
>>> diff --git a/lib/e2p/feature.c b/lib/e2p/feature.c
>>> index 16fba53..965fc16 100644
>>> --- a/lib/e2p/feature.c
>>> +++ b/lib/e2p/feature.c
>>> @@ -40,6 +40,8 @@ static struct feature feature_list[] = {
>>> "resize_inode" },
>>> { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_LAZY_BG,
>>> "lazy_bg" },
>>> + { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP,
>>> + "snapshot" },
>>>
>>> { E2P_FEATURE_RO_INCOMPAT, EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER,
>>> "sparse_super" },
>>> @@ -59,6 +61,8 @@ static struct feature feature_list[] = {
>>> "quota" },
>>> { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_BIGALLOC,
>>> "bigalloc"},
>>> + { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_METADATA_CSUM,
>>> + "metadata_csum"},
>>>
>>> { E2P_FEATURE_INCOMPAT, EXT2_FEATURE_INCOMPAT_COMPRESSION,
>>> "compression" },
>>> diff --git a/lib/e2p/ls.c b/lib/e2p/ls.c
>>> index 0f36f40..aaacdaa 100644
>>> --- a/lib/e2p/ls.c
>>> +++ b/lib/e2p/ls.c
>>> @@ -413,6 +413,10 @@ void list_super2(struct ext2_super_block * sb, FILE *f)
>>> if (sb->s_grp_quota_inum)
>>> fprintf(f, "Group quota inode: %u\n",
>>> sb->s_grp_quota_inum);
>>> +
>>> + if (sb->s_feature_ro_compat & EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)
>>> + fprintf(f, "Checksum: 0x%08x\n",
>>> + sb->s_checksum);
>>> }
>>>
>>> void list_super (struct ext2_super_block * s)
>>> diff --git a/lib/ext2fs/ext2_fs.h b/lib/ext2fs/ext2_fs.h
>>> index 4fec5db..1b02054 100644
>>> --- a/lib/ext2fs/ext2_fs.h
>>> +++ b/lib/ext2fs/ext2_fs.h
>>> @@ -142,7 +142,9 @@ struct ext2_group_desc
>>> __u16 bg_free_inodes_count; /* Free inodes count */
>>> __u16 bg_used_dirs_count; /* Directories count */
>>> __u16 bg_flags;
>>> - __u32 bg_reserved[2];
>>> + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
>>> + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
>>> + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
>>> __u16 bg_itable_unused; /* Unused inodes count */
>>> __u16 bg_checksum; /* crc16(s_uuid+grouo_num+group_desc)*/
>>> };
>>> @@ -159,7 +161,9 @@ struct ext4_group_desc
>>> __u16 bg_free_inodes_count; /* Free inodes count */
>>> __u16 bg_used_dirs_count; /* Directories count */
>>> __u16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
>>> - __u32 bg_reserved[2]; /* Likely block/inode bitmap checksum */
>>> + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
>>> + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
>>> + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
>>> __u16 bg_itable_unused; /* Unused inodes count */
>>> __u16 bg_checksum; /* crc16(sb_uuid+group+desc) */
>>> __u32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
>>> @@ -169,7 +173,10 @@ struct ext4_group_desc
>>> __u16 bg_free_inodes_count_hi;/* Free inodes count MSB */
>>> __u16 bg_used_dirs_count_hi; /* Directories count MSB */
>>> __u16 bg_itable_unused_hi; /* Unused inodes count MSB */
>>> - __u32 bg_reserved2[3];
>>> + __u32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
>>> + __u16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
>>> + __u16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
>>> + __u32 bg_reserved;
>>> };
>>>
>>> #define EXT2_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not initialized */
>>> @@ -363,7 +370,7 @@ struct ext2_inode {
>>> __u16 l_i_file_acl_high;
>>> __u16 l_i_uid_high; /* these 2 fields */
>>> __u16 l_i_gid_high; /* were reserved2[0] */
>>> - __u32 l_i_reserved2;
>>> + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
>>> } linux2;
>>> struct {
>>> __u8 h_i_frag; /* Fragment number */
>>> @@ -410,7 +417,7 @@ struct ext2_inode_large {
>>> __u16 l_i_file_acl_high;
>>> __u16 l_i_uid_high; /* these 2 fields */
>>> __u16 l_i_gid_high; /* were reserved2[0] */
>>> - __u32 l_i_reserved2;
>>> + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
>>> } linux2;
>>> struct {
>>> __u8 h_i_frag; /* Fragment number */
>>> @@ -441,7 +448,7 @@ struct ext2_inode_large {
>>> #define i_gid_low i_gid
>>> #define i_uid_high osd2.linux2.l_i_uid_high
>>> #define i_gid_high osd2.linux2.l_i_gid_high
>>> -#define i_reserved2 osd2.linux2.l_i_reserved2
>>> +#define i_checksum osd2.linux2.l_i_checksum
>>> #else
>>> #if defined(__GNU__)
>>>
>>> @@ -623,7 +630,8 @@ struct ext2_super_block {
>>> __u32 s_usr_quota_inum; /* inode number of user quota file */
>>> __u32 s_grp_quota_inum; /* inode number of group quota file */
>>> __u32 s_overhead_blocks; /* overhead blocks/clusters in fs */
>>> - __u32 s_reserved[109]; /* Padding to the end of the block */
>>> + __u32 s_checksum; /* crc32c(superblock) */
>>> + __u32 s_reserved[108]; /* Padding to the end of the block */
>>> };
>>>
>>> #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
>>> @@ -671,7 +679,9 @@ struct ext2_super_block {
>>> #define EXT2_FEATURE_COMPAT_RESIZE_INODE 0x0010
>>> #define EXT2_FEATURE_COMPAT_DIR_INDEX 0x0020
>>> #define EXT2_FEATURE_COMPAT_LAZY_BG 0x0040
>>> -#define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080
>>> +/* #define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080 not used, legacy */
>>> +#define EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP 0x0100
>>> +
>>>
>>> #define EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
>>> #define EXT2_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
>>> @@ -683,6 +693,7 @@ struct ext2_super_block {
>>> #define EXT4_FEATURE_RO_COMPAT_HAS_SNAPSHOT 0x0080
>>> #define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
>>> #define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
>>> +#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
>>>
>>> #define EXT2_FEATURE_INCOMPAT_COMPRESSION 0x0001
>>> #define EXT2_FEATURE_INCOMPAT_FILETYPE 0x0002
>>> diff --git a/lib/ext2fs/swapfs.c b/lib/ext2fs/swapfs.c
>>> index 87b1a2e..d1c4a56 100644
>>> --- a/lib/ext2fs/swapfs.c
>>> +++ b/lib/ext2fs/swapfs.c
>>> @@ -78,6 +78,8 @@ void ext2fs_swap_super(struct ext2_super_block * sb)
>>> sb->s_snapshot_list = ext2fs_swab32(sb->s_snapshot_list);
>>> sb->s_usr_quota_inum = ext2fs_swab32(sb->s_usr_quota_inum);
>>> sb->s_grp_quota_inum = ext2fs_swab32(sb->s_grp_quota_inum);
>>> + sb->s_overhead_blocks = ext2fs_swab32(sb->s_overhead_blocks);
>>> + sb->s_checksum = ext2fs_swab32(sb->s_checksum);
>>>
>>> for (i=0; i < 4; i++)
>>> sb->s_hash_seed[i] = ext2fs_swab32(sb->s_hash_seed[i]);
>>> @@ -106,6 +108,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
>>> gdp->bg_free_inodes_count = ext2fs_swab16(gdp->bg_free_inodes_count);
>>> gdp->bg_used_dirs_count = ext2fs_swab16(gdp->bg_used_dirs_count);
>>> gdp->bg_flags = ext2fs_swab16(gdp->bg_flags);
>>> + gdp->bg_exclude_bitmap_lo = ext2fs_swab32(gdp->bg_exclude_bitmap_lo);
>>> + gdp->bg_block_bitmap_csum_lo =
>>> + ext2fs_swab16(gdp->bg_block_bitmap_csum_lo);
>>> + gdp->bg_inode_bitmap_csum_lo =
>>> + ext2fs_swab16(gdp->bg_inode_bitmap_csum_lo);
>>> gdp->bg_itable_unused = ext2fs_swab16(gdp->bg_itable_unused);
>>> gdp->bg_checksum = ext2fs_swab16(gdp->bg_checksum);
>>> /* If we're 32-bit, we're done */
>>> @@ -125,6 +132,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
>>> gdp4->bg_used_dirs_count_hi =
>>> ext2fs_swab16(gdp4->bg_used_dirs_count_hi);
>>> gdp4->bg_itable_unused_hi = ext2fs_swab16(gdp4->bg_itable_unused_hi);
>>> + gdp->bg_exclude_bitmap_hi = ext2fs_swab16(gdp->bg_exclude_bitmap_hi);
>>> + gdp->bg_block_bitmap_csum_hi =
>>> + ext2fs_swab16(gdp->bg_block_bitmap_csum_hi);
>>> + gdp->bg_inode_bitmap_csum_hi =
>>> + ext2fs_swab16(gdp->bg_inode_bitmap_csum_hi);
>>> }
>>>
>>> void ext2fs_swap_group_desc(struct ext2_group_desc *gdp)
>>> @@ -244,8 +256,8 @@ void ext2fs_swap_inode_full(ext2_filsys fs, struct ext2_inode_large *t,
>>> ext2fs_swab16 (f->osd2.linux2.l_i_uid_high);
>>> t->osd2.linux2.l_i_gid_high =
>>> ext2fs_swab16 (f->osd2.linux2.l_i_gid_high);
>>> - t->osd2.linux2.l_i_reserved2 =
>>> - ext2fs_swab32(f->osd2.linux2.l_i_reserved2);
>>> + t->osd2.linux2.l_i_checksum =
>>> + ext2fs_swab32(f->osd2.linux2.checksum);
>>> break;
>>> case EXT2_OS_HURD:
>>> t->osd1.hurd1.h_i_translator =
>>> diff --git a/lib/ext2fs/tst_inode_size.c b/lib/ext2fs/tst_inode_size.c
>>> index 962f1cd..683b79c 100644
>>> --- a/lib/ext2fs/tst_inode_size.c
>>> +++ b/lib/ext2fs/tst_inode_size.c
>>> @@ -61,7 +61,7 @@ void check_structure_fields()
>>> check_field(osd2.linux2.l_i_file_acl_high);
>>> check_field(osd2.linux2.l_i_uid_high);
>>> check_field(osd2.linux2.l_i_gid_high);
>>> - check_field(osd2.linux2.l_i_reserved2);
>>> + check_field(osd2.linux2.l_i_checksum);
>>> printf("Ending offset is %d\n\n", cur_offset);
>>> #endif
>>> }
>>> diff --git a/lib/ext2fs/tst_super_size.c b/lib/ext2fs/tst_super_size.c
>>> index 1e5a524..75659ae 100644
>>> --- a/lib/ext2fs/tst_super_size.c
>>> +++ b/lib/ext2fs/tst_super_size.c
>>> @@ -126,6 +126,7 @@ void check_superblock_fields()
>>> check_field(s_usr_quota_inum);
>>> check_field(s_grp_quota_inum);
>>> check_field(s_overhead_blocks);
>>> + check_field(s_checksum);
>>> check_field(s_reserved);
>>> printf("Ending offset is %d\n\n", cur_offset);
>>> #endif
>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
Cheers, Andreas
On Thu, Sep 15, 2011 at 01:10:41PM -0600, Andreas Dilger wrote:
>
> This is also a possible alternative, though it makes for more fragments that
> need to be checksummed. I think as a general rule it makes sense to store
> the checksum as the last word in the structure, if possible, so that the
> checksum can be computed in a single call. This is already done for 128-byte
> inodes and for 32-byte group descriptors, but should also be done for the
> s_checksum field in the superblock (i.e. put it after s_reserved instead of
> before).
Or we just zero out the checksum field, checksum the entire data
structure, and then fill in the newly calculated checksum. In fact
that's what I assumed Darrick was going to do.
- Ted
On Thu, Sep 15, 2011 at 01:10:41PM -0600, Andreas Dilger wrote:
> On 2011-09-15, at 11:19 AM, Darrick J. Wong wrote:
> > On Thu, Sep 15, 2011 at 09:55:12AM -0700, Darrick J. Wong wrote:
> >> On Thu, Sep 15, 2011 at 11:12:22AM -0400, Theodore Ts'o wrote:
> >>>
> >>> Hi Darrick, Amir,
> >>>
> >>> Could you please take a look at the changes here and make sure they look
> >>> sane to you? I just want to make sure we're all on the same page as
> >>> far as on-disk field assignments are concerned.
> >>>
> >>> Also, one thought for Darrick. I note that with the assignment of
> >>> l_i_checksum, we've exhausted the very last field in the base 128-byte
> >>> inode. Given that the checksum is protecting a 128 byte or 256 byte
> >>> inode, I wonder if we need to use a full 32 bit checksum. Maybe we
> >>> should just use 16 bits of a crc32c?
> >>>
> >>> I did some web searching on the issue, and the recommendation I've come
> >>> across is the following:
> >>>
> >>> "Generally speaking, an n-bit CRC's error detection properties
> >>> degrade after 2**(n-1)-1 data bits."
> >>>
> >>> So a CRC-16 will be good up to just under 4k. A 16-bit truncated crc32c
> >>> will presumably not be as good as a crc16, but seems to me that it's
> >>> probably fine for a 128-256 byte inode. Especially since the main thing
> >>> we're generally worried about is detecting a block getting written to
> >>> the wrong location, overwriting an existing inode table block. So if we
> >>> were really paranoid we could verify the checksums for all of the inodes
> >>> in a particular inode table block when we read in the inode table block
> >>> in question.
> >>
> >> On the other hand, you can set inode_size = block_size, which means that
> >> with a 4k inode + 32-bit inode number + 16-byte UUID you actually could
> >> run afoul of that degradation. But that seems like an extreme argument
> >> for an infrequent case.
> >>
> >> Actually, I've started wondering if we could split the 4 bytes of the crc32c
> >> among the first few inodes of the block, and compute the checksums at block
> >> size granularity. Though that would make inode updates particularly more
> >> expensive... but if I'm going to shift the write-time checksum to a journal
> >> callback then it's not going to matter (for the journal-using users, anyway).
> >>
> >> Though with that scheme, you'd probably lose more inodes for any given
> >> integrity error. It also means that the checksum size in each inode becomes
> >> variable (32 bits if inode=blocksize, 16 if inode=blocksize/2, and 8
> >> otherwise), which is a somewhat confusing schema.
> >>
> >> <shrug> Do you anticipate a need to add more fields to 128-byte inode
> >> filesystems? I think most of those would be former ext2/3 filesystems,
> >> floppies, and "small" filesystems, correct?
> >>
> >> Or does this second scheme sound more attractive?
> >
> > I forgot to say, "as opposed to storing the lower 16 bits below 128 bytes and
> > the upper 16 bits somewhere above it."
>
> This is also a possible alternative, though it makes for more fragments that
> need to be checksummed. I think as a general rule it makes sense to store
> the checksum as the last word in the structure, if possible, so that the
> checksum can be computed in a single call. This is already done for 128-byte
> inodes and for 32-byte group descriptors, but should also be done for the
> s_checksum field in the superblock (i.e. put it after s_reserved instead of
> before).
>
> For inodes 256-bytes or larger (which are commonly used for Lustre to store
> large xattrs that are needed for every file access), and 64-byte group
> descriptors it has to checksum 2 fragments, but at least not more than that
> if we precompute for each inode the crc32c(uuid + inum) seed and for each
> group the crc32c(uuid + group) seed. The superblock already contains the
> UUID, but it may make sense to still precompute the crc32c(uuid) part and
> store it in ext4_sb_info for computing the inode seed.
>
> It really would be interesting to measure the crc32c() and crc16() performance
> for 512MB in chunks of 4, 32, 128, 256, and 4096 bytes (which is the largest
> that we will generally use until we get to data checksums). That would give
> us a good idea how fast the checksums _really_ are in our actual usage.
Ok, here's a rough cut with the Xeon X5650 at work:
crc32c-sby8-be@4: sz=536870912 time=1816.711ms speed=288591.85K/s res=0xdc9b4e3f
crc32c-sby8-be@32: sz=536870912 time=419.187ms speed=1250724.37K/s res=0xdc9b4e3f
crc32c-sby8-be@128: sz=536870912 time=340.387ms speed=1540270.88K/s res=0xdc9b4e3f
crc32c-sby8-be@256: sz=536870912 time=327.602ms speed=1600381.50K/s res=0xdc9b4e3f
crc32c-sby8-be@512: sz=536870912 time=321.288ms speed=1631831.48K/s res=0xdc9b4e3f
crc32c-sby8-be@4096: sz=536870912 time=314.795ms speed=1665489.62K/s res=0xdc9b4e3f
crc32c-sby8-be@65536: sz=536870912 time=310.732ms speed=1687266.98K/s res=0xdc9b4e3f
crc32c-sby8-be@1048576: sz=536870912 time=310.512ms speed=1688461.53K/s res=0xdc9b4e3f
crc32c-sby8-be@536870912: sz=536870912 time=312.628ms speed=1677032.72K/s res=0xdc9b4e3f
crc32c-sby8-le@4: sz=536870912 time=1722.019ms speed=304461.26K/s res=0xfc72d93b
crc32c-sby8-le@32: sz=536870912 time=410.747ms speed=1276424.81K/s res=0xfc72d93b
crc32c-sby8-le@128: sz=536870912 time=338.527ms speed=1548731.75K/s res=0xfc72d93b
crc32c-sby8-le@256: sz=536870912 time=329.869ms speed=1589382.17K/s res=0xfc72d93b
crc32c-sby8-le@512: sz=536870912 time=322.203ms speed=1627196.09K/s res=0xfc72d93b
crc32c-sby8-le@4096: sz=536870912 time=309.987ms speed=1691324.57K/s res=0xfc72d93b
crc32c-sby8-le@65536: sz=536870912 time=309.072ms speed=1696328.93K/s res=0xfc72d93b
crc32c-sby8-le@1048576: sz=536870912 time=310.918ms speed=1686259.03K/s res=0xfc72d93b
crc32c-sby8-le@536870912: sz=536870912 time=309.269ms speed=1695251.44K/s res=0xfc72d93b
crc16@4: sz=536870912 time=1605.590ms speed=326539.15K/s res=0x4bff
crc16@32: sz=536870912 time=1440.110ms speed=364061.15K/s res=0x4bff
crc16@128: sz=536870912 time=1374.726ms speed=381376.47K/s res=0x4bff
crc16@256: sz=536870912 time=1375.174ms speed=381251.98K/s res=0x4bff
crc16@512: sz=536870912 time=1393.672ms speed=376191.85K/s res=0x4bff
crc16@4096: sz=536870912 time=1365.874ms speed=383847.97K/s res=0x4bff
crc16@65536: sz=536870912 time=1363.806ms speed=384430.11K/s res=0x4bff
crc16@1048576: sz=536870912 time=1363.767ms speed=384441.16K/s res=0x4bff
crc16@536870912: sz=536870912 time=1366.145ms speed=383771.74K/s res=0x4bff
crc32c@4: sz=536870912 time=1435.798ms speed=365154.39K/s res=0xfc72d93b
crc32c@32: sz=536870912 time=1287.815ms speed=407114.42K/s res=0xfc72d93b
crc32c@128: sz=536870912 time=1249.553ms speed=419580.42K/s res=0xfc72d93b
crc32c@256: sz=536870912 time=1244.488ms speed=421288.07K/s res=0xfc72d93b
crc32c@512: sz=536870912 time=1240.666ms speed=422585.95K/s res=0xfc72d93b
crc32c@4096: sz=536870912 time=1238.181ms speed=423434.03K/s res=0xfc72d93b
crc32c@65536: sz=536870912 time=1238.595ms speed=423292.52K/s res=0xfc72d93b
crc32c@1048576: sz=536870912 time=1239.664ms speed=422927.60K/s res=0xfc72d93b
crc32c@536870912: sz=536870912 time=1238.147ms speed=423445.61K/s res=0xfc72d93b
crc16-t10dif@4: sz=536870912 time=1898.668ms speed=276134.71K/s res=0x7449
crc16-t10dif@32: sz=536870912 time=1781.657ms speed=294269.88K/s res=0x7449
crc16-t10dif@128: sz=536870912 time=1768.654ms speed=296433.30K/s res=0x7449
crc16-t10dif@256: sz=536870912 time=1767.750ms speed=296584.86K/s res=0x7449
crc16-t10dif@512: sz=536870912 time=1766.331ms speed=296823.21K/s res=0x7449
crc16-t10dif@4096: sz=536870912 time=1765.198ms speed=297013.75K/s res=0x7449
crc16-t10dif@65536: sz=536870912 time=1764.398ms speed=297148.38K/s res=0x7449
crc16-t10dif@1048576: sz=536870912 time=1765.903ms speed=296895.06K/s res=0x7449
crc16-t10dif@536870912: sz=536870912 time=1765.477ms speed=296966.71K/s res=0x7449
crc32c-intel@4: sz=536870912 time=1025.192ms speed=511404.47K/s res=0xfc72d93b
crc32c-intel@32: sz=536870912 time=135.124ms speed=3880043.59K/s res=0xfc72d93b
crc32c-intel@128: sz=536870912 time=121.991ms speed=4297766.44K/s res=0xfc72d93b
crc32c-intel@256: sz=536870912 time=120.008ms speed=4368783.42K/s res=0xfc72d93b
crc32c-intel@512: sz=536870912 time=118.889ms speed=4409903.92K/s res=0xfc72d93b
crc32c-intel@4096: sz=536870912 time=118.112ms speed=4438891.54K/s res=0xfc72d93b
crc32c-intel@65536: sz=536870912 time=118.010ms speed=4442735.48K/s res=0xfc72d93b
crc32c-intel@1048576: sz=536870912 time=119.969ms speed=4370205.80K/s res=0xfc72d93b
crc32c-intel@536870912: sz=536870912 time=118.369ms speed=4429255.67K/s res=0xfc72d93b
As you can see from the results, the algorithm(s) that are fast generally don't
reach full speed until they hit 4KB chunk sizes. powerpc64 and a laptop seems
to yield speed scaling similar results.
crc32c-sby8-* = my new crc32c implementation
crc16 = kernel's crc16 implementation
crc32c = kernel's current crc32c sw implementation
crc16-t10dif = t10dif crc16 implementation
crc32c-intel = hw accelerated crc32c
--D
>
> >>> commit ceade753f14f2697d329f71b5277b49fd46fcb55
> >>> Author: Theodore Ts'o <[email protected]>
> >>> Date: Thu Sep 15 10:38:55 2011 -0400
> >>>
> >>> libext2fs: add metadata checksum and snapshot feature flags
> >>>
> >>> Reserve EXT4_FEATURE_RO_COMPAT_METADATA_CSUM and
> >>> EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP. Also reserve fields in the
> >>> superblock and the inode for the checksums. In the block group
> >>> descriptor, reserve the exclude bitmap field for the snapshot feature,
> >>> and checksums for the inode and block allocation bitmaps.
> >>>
> >>> With this commit, the metadata checksum and exclude bitmap features
> >>> should have reserved all of the fields they need in ext4's on-disk
> >>> format.
> >>>
> >>> This commit also fixes an a missing byte swap for s_overhead_blocks.
> >>>
> >>> Signed-off-by: "Theodore Ts'o" <[email protected]>
> >>> Cc: Darrick J. Wong <[email protected]>
> >>> Cc: Amir Goldstein <[email protected]>
> >>>
> >>> diff --git a/debugfs/set_fields.c b/debugfs/set_fields.c
> >>> index ac6bc25..cba9c12 100644
> >>> --- a/debugfs/set_fields.c
> >>> +++ b/debugfs/set_fields.c
> >>> @@ -144,6 +144,7 @@ static struct field_set_info super_fields[] = {
> >>> { "usr_quota_inum", &set_sb.s_usr_quota_inum, 4, parse_uint },
> >>> { "grp_quota_inum", &set_sb.s_grp_quota_inum, 4, parse_uint },
> >>> { "overhead_blocks", &set_sb.s_overhead_blocks, 4, parse_uint },
> >>> + { "checksum", &set_sb.s_checksum, 2, parse_uint },
> >>> { 0, 0, 0, 0 }
> >>> };
> >>>
> >>> @@ -179,6 +180,7 @@ static struct field_set_info inode_fields[] = {
> >>> { "fsize", &set_inode.osd2.hurd2.h_i_fsize, 1, parse_uint },
> >>> { "uid_high", &set_inode.osd2.linux2.l_i_uid_high, 2, parse_uint },
> >>> { "gid_high", &set_inode.osd2.linux2.l_i_gid_high, 2, parse_uint },
> >>> + { "checksum", &set_inode.osd2.linux2.l_i_checksum, 4, parse_uint },
> >>> { "author", &set_inode.osd2.hurd2.h_i_author, 4, parse_uint },
> >>> { "bmap", NULL, 4, parse_bmap, FLAG_ARRAY },
> >>> { 0, 0, 0, 0 }
> >>> @@ -192,7 +194,6 @@ static struct field_set_info ext2_bg_fields[] = {
> >>> { "free_inodes_count", &set_gd.bg_free_inodes_count, 2, parse_uint },
> >>> { "used_dirs_count", &set_gd.bg_used_dirs_count, 2, parse_uint },
> >>> { "flags", &set_gd.bg_flags, 2, parse_uint },
> >>> - { "reserved", &set_gd.bg_reserved, 2, parse_uint, FLAG_ARRAY, 2 },
> >>> { "itable_unused", &set_gd.bg_itable_unused, 2, parse_uint },
> >>> { "checksum", &set_gd.bg_checksum, 2, parse_gd_csum },
> >>> { 0, 0, 0, 0 }
> >>> diff --git a/lib/e2p/feature.c b/lib/e2p/feature.c
> >>> index 16fba53..965fc16 100644
> >>> --- a/lib/e2p/feature.c
> >>> +++ b/lib/e2p/feature.c
> >>> @@ -40,6 +40,8 @@ static struct feature feature_list[] = {
> >>> "resize_inode" },
> >>> { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_LAZY_BG,
> >>> "lazy_bg" },
> >>> + { E2P_FEATURE_COMPAT, EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP,
> >>> + "snapshot" },
> >>>
> >>> { E2P_FEATURE_RO_INCOMPAT, EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER,
> >>> "sparse_super" },
> >>> @@ -59,6 +61,8 @@ static struct feature feature_list[] = {
> >>> "quota" },
> >>> { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_BIGALLOC,
> >>> "bigalloc"},
> >>> + { E2P_FEATURE_RO_INCOMPAT, EXT4_FEATURE_RO_COMPAT_METADATA_CSUM,
> >>> + "metadata_csum"},
> >>>
> >>> { E2P_FEATURE_INCOMPAT, EXT2_FEATURE_INCOMPAT_COMPRESSION,
> >>> "compression" },
> >>> diff --git a/lib/e2p/ls.c b/lib/e2p/ls.c
> >>> index 0f36f40..aaacdaa 100644
> >>> --- a/lib/e2p/ls.c
> >>> +++ b/lib/e2p/ls.c
> >>> @@ -413,6 +413,10 @@ void list_super2(struct ext2_super_block * sb, FILE *f)
> >>> if (sb->s_grp_quota_inum)
> >>> fprintf(f, "Group quota inode: %u\n",
> >>> sb->s_grp_quota_inum);
> >>> +
> >>> + if (sb->s_feature_ro_compat & EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)
> >>> + fprintf(f, "Checksum: 0x%08x\n",
> >>> + sb->s_checksum);
> >>> }
> >>>
> >>> void list_super (struct ext2_super_block * s)
> >>> diff --git a/lib/ext2fs/ext2_fs.h b/lib/ext2fs/ext2_fs.h
> >>> index 4fec5db..1b02054 100644
> >>> --- a/lib/ext2fs/ext2_fs.h
> >>> +++ b/lib/ext2fs/ext2_fs.h
> >>> @@ -142,7 +142,9 @@ struct ext2_group_desc
> >>> __u16 bg_free_inodes_count; /* Free inodes count */
> >>> __u16 bg_used_dirs_count; /* Directories count */
> >>> __u16 bg_flags;
> >>> - __u32 bg_reserved[2];
> >>> + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
> >>> + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> >>> + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> >>> __u16 bg_itable_unused; /* Unused inodes count */
> >>> __u16 bg_checksum; /* crc16(s_uuid+grouo_num+group_desc)*/
> >>> };
> >>> @@ -159,7 +161,9 @@ struct ext4_group_desc
> >>> __u16 bg_free_inodes_count; /* Free inodes count */
> >>> __u16 bg_used_dirs_count; /* Directories count */
> >>> __u16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
> >>> - __u32 bg_reserved[2]; /* Likely block/inode bitmap checksum */
> >>> + __u32 bg_exclude_bitmap_lo; /* Exclude bitmap for snapshots */
> >>> + __u16 bg_block_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> >>> + __u16 bg_inode_bitmap_csum_lo;/* crc32c(s_uuid+grp_num+bitmap) LSB */
> >>> __u16 bg_itable_unused; /* Unused inodes count */
> >>> __u16 bg_checksum; /* crc16(sb_uuid+group+desc) */
> >>> __u32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
> >>> @@ -169,7 +173,10 @@ struct ext4_group_desc
> >>> __u16 bg_free_inodes_count_hi;/* Free inodes count MSB */
> >>> __u16 bg_used_dirs_count_hi; /* Directories count MSB */
> >>> __u16 bg_itable_unused_hi; /* Unused inodes count MSB */
> >>> - __u32 bg_reserved2[3];
> >>> + __u32 bg_exclude_bitmap_hi; /* Exclude bitmap block MSB */
> >>> + __u16 bg_block_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
> >>> + __u16 bg_inode_bitmap_csum_hi;/* crc32c(s_uuid+grp_num+bitmap) MSB */
> >>> + __u32 bg_reserved;
> >>> };
> >>>
> >>> #define EXT2_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not initialized */
> >>> @@ -363,7 +370,7 @@ struct ext2_inode {
> >>> __u16 l_i_file_acl_high;
> >>> __u16 l_i_uid_high; /* these 2 fields */
> >>> __u16 l_i_gid_high; /* were reserved2[0] */
> >>> - __u32 l_i_reserved2;
> >>> + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
> >>> } linux2;
> >>> struct {
> >>> __u8 h_i_frag; /* Fragment number */
> >>> @@ -410,7 +417,7 @@ struct ext2_inode_large {
> >>> __u16 l_i_file_acl_high;
> >>> __u16 l_i_uid_high; /* these 2 fields */
> >>> __u16 l_i_gid_high; /* were reserved2[0] */
> >>> - __u32 l_i_reserved2;
> >>> + __u32 l_i_checksum; /* crc32c(uuid+inum+inode) */
> >>> } linux2;
> >>> struct {
> >>> __u8 h_i_frag; /* Fragment number */
> >>> @@ -441,7 +448,7 @@ struct ext2_inode_large {
> >>> #define i_gid_low i_gid
> >>> #define i_uid_high osd2.linux2.l_i_uid_high
> >>> #define i_gid_high osd2.linux2.l_i_gid_high
> >>> -#define i_reserved2 osd2.linux2.l_i_reserved2
> >>> +#define i_checksum osd2.linux2.l_i_checksum
> >>> #else
> >>> #if defined(__GNU__)
> >>>
> >>> @@ -623,7 +630,8 @@ struct ext2_super_block {
> >>> __u32 s_usr_quota_inum; /* inode number of user quota file */
> >>> __u32 s_grp_quota_inum; /* inode number of group quota file */
> >>> __u32 s_overhead_blocks; /* overhead blocks/clusters in fs */
> >>> - __u32 s_reserved[109]; /* Padding to the end of the block */
> >>> + __u32 s_checksum; /* crc32c(superblock) */
> >>> + __u32 s_reserved[108]; /* Padding to the end of the block */
> >>> };
> >>>
> >>> #define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
> >>> @@ -671,7 +679,9 @@ struct ext2_super_block {
> >>> #define EXT2_FEATURE_COMPAT_RESIZE_INODE 0x0010
> >>> #define EXT2_FEATURE_COMPAT_DIR_INDEX 0x0020
> >>> #define EXT2_FEATURE_COMPAT_LAZY_BG 0x0040
> >>> -#define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080
> >>> +/* #define EXT2_FEATURE_COMPAT_EXCLUDE_INODE 0x0080 not used, legacy */
> >>> +#define EXT2_FEATURE_COMPAT_EXCLUDE_BITMAP 0x0100
> >>> +
> >>>
> >>> #define EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
> >>> #define EXT2_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
> >>> @@ -683,6 +693,7 @@ struct ext2_super_block {
> >>> #define EXT4_FEATURE_RO_COMPAT_HAS_SNAPSHOT 0x0080
> >>> #define EXT4_FEATURE_RO_COMPAT_QUOTA 0x0100
> >>> #define EXT4_FEATURE_RO_COMPAT_BIGALLOC 0x0200
> >>> +#define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM 0x0400
> >>>
> >>> #define EXT2_FEATURE_INCOMPAT_COMPRESSION 0x0001
> >>> #define EXT2_FEATURE_INCOMPAT_FILETYPE 0x0002
> >>> diff --git a/lib/ext2fs/swapfs.c b/lib/ext2fs/swapfs.c
> >>> index 87b1a2e..d1c4a56 100644
> >>> --- a/lib/ext2fs/swapfs.c
> >>> +++ b/lib/ext2fs/swapfs.c
> >>> @@ -78,6 +78,8 @@ void ext2fs_swap_super(struct ext2_super_block * sb)
> >>> sb->s_snapshot_list = ext2fs_swab32(sb->s_snapshot_list);
> >>> sb->s_usr_quota_inum = ext2fs_swab32(sb->s_usr_quota_inum);
> >>> sb->s_grp_quota_inum = ext2fs_swab32(sb->s_grp_quota_inum);
> >>> + sb->s_overhead_blocks = ext2fs_swab32(sb->s_overhead_blocks);
> >>> + sb->s_checksum = ext2fs_swab32(sb->s_checksum);
> >>>
> >>> for (i=0; i < 4; i++)
> >>> sb->s_hash_seed[i] = ext2fs_swab32(sb->s_hash_seed[i]);
> >>> @@ -106,6 +108,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
> >>> gdp->bg_free_inodes_count = ext2fs_swab16(gdp->bg_free_inodes_count);
> >>> gdp->bg_used_dirs_count = ext2fs_swab16(gdp->bg_used_dirs_count);
> >>> gdp->bg_flags = ext2fs_swab16(gdp->bg_flags);
> >>> + gdp->bg_exclude_bitmap_lo = ext2fs_swab32(gdp->bg_exclude_bitmap_lo);
> >>> + gdp->bg_block_bitmap_csum_lo =
> >>> + ext2fs_swab16(gdp->bg_block_bitmap_csum_lo);
> >>> + gdp->bg_inode_bitmap_csum_lo =
> >>> + ext2fs_swab16(gdp->bg_inode_bitmap_csum_lo);
> >>> gdp->bg_itable_unused = ext2fs_swab16(gdp->bg_itable_unused);
> >>> gdp->bg_checksum = ext2fs_swab16(gdp->bg_checksum);
> >>> /* If we're 32-bit, we're done */
> >>> @@ -125,6 +132,11 @@ void ext2fs_swap_group_desc2(ext2_filsys fs, struct ext2_group_desc *gdp)
> >>> gdp4->bg_used_dirs_count_hi =
> >>> ext2fs_swab16(gdp4->bg_used_dirs_count_hi);
> >>> gdp4->bg_itable_unused_hi = ext2fs_swab16(gdp4->bg_itable_unused_hi);
> >>> + gdp->bg_exclude_bitmap_hi = ext2fs_swab16(gdp->bg_exclude_bitmap_hi);
> >>> + gdp->bg_block_bitmap_csum_hi =
> >>> + ext2fs_swab16(gdp->bg_block_bitmap_csum_hi);
> >>> + gdp->bg_inode_bitmap_csum_hi =
> >>> + ext2fs_swab16(gdp->bg_inode_bitmap_csum_hi);
> >>> }
> >>>
> >>> void ext2fs_swap_group_desc(struct ext2_group_desc *gdp)
> >>> @@ -244,8 +256,8 @@ void ext2fs_swap_inode_full(ext2_filsys fs, struct ext2_inode_large *t,
> >>> ext2fs_swab16 (f->osd2.linux2.l_i_uid_high);
> >>> t->osd2.linux2.l_i_gid_high =
> >>> ext2fs_swab16 (f->osd2.linux2.l_i_gid_high);
> >>> - t->osd2.linux2.l_i_reserved2 =
> >>> - ext2fs_swab32(f->osd2.linux2.l_i_reserved2);
> >>> + t->osd2.linux2.l_i_checksum =
> >>> + ext2fs_swab32(f->osd2.linux2.checksum);
> >>> break;
> >>> case EXT2_OS_HURD:
> >>> t->osd1.hurd1.h_i_translator =
> >>> diff --git a/lib/ext2fs/tst_inode_size.c b/lib/ext2fs/tst_inode_size.c
> >>> index 962f1cd..683b79c 100644
> >>> --- a/lib/ext2fs/tst_inode_size.c
> >>> +++ b/lib/ext2fs/tst_inode_size.c
> >>> @@ -61,7 +61,7 @@ void check_structure_fields()
> >>> check_field(osd2.linux2.l_i_file_acl_high);
> >>> check_field(osd2.linux2.l_i_uid_high);
> >>> check_field(osd2.linux2.l_i_gid_high);
> >>> - check_field(osd2.linux2.l_i_reserved2);
> >>> + check_field(osd2.linux2.l_i_checksum);
> >>> printf("Ending offset is %d\n\n", cur_offset);
> >>> #endif
> >>> }
> >>> diff --git a/lib/ext2fs/tst_super_size.c b/lib/ext2fs/tst_super_size.c
> >>> index 1e5a524..75659ae 100644
> >>> --- a/lib/ext2fs/tst_super_size.c
> >>> +++ b/lib/ext2fs/tst_super_size.c
> >>> @@ -126,6 +126,7 @@ void check_superblock_fields()
> >>> check_field(s_usr_quota_inum);
> >>> check_field(s_grp_quota_inum);
> >>> check_field(s_overhead_blocks);
> >>> + check_field(s_checksum);
> >>> check_field(s_reserved);
> >>> printf("Ending offset is %d\n\n", cur_offset);
> >>> #endif
> >>>
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> >> the body of a message to [email protected]
> >> More majordomo info at http://vger.kernel.org/majordomo-info.html
> >>
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> > the body of a message to [email protected]
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
>
>
> Cheers, Andreas
>
>
>
>
>
On 2011-09-15, at 2:05 PM, Darrick J. Wong wrote:
> On Thu, Sep 15, 2011 at 01:10:41PM -0600, Andreas Dilger wrote:
>>
>> It really would be interesting to measure the crc32c() and crc16() performance
>> for 512MB in chunks of 4, 32, 128, 256, and 4096 bytes (which is the largest
>> that we will generally use until we get to data checksums). That would give
>> us a good idea how fast the checksums _really_ are in our actual usage.
>
> Ok, here's a rough cut with the Xeon X5650 at work:
>
> crc32c-sby8-le@4: sz=536870912 time=1722.019ms speed=304461.26K/s
> crc32c-sby8-le@32: sz=536870912 time=410.747ms speed=1276424.81K/s
> crc32c-sby8-le@128: sz=536870912 time=338.527ms speed=1548731.75K/s
> crc32c-sby8-le@256: sz=536870912 time=329.869ms speed=1589382.17K/s
:
:
> crc32c-sby8-le@536870912: sz=536870912 time=309.269ms speed=1695251.44K/s
This gets 75% peak speed at 32 bytes, and is 91% of the peak at 128 bytes.
> crc16@4: sz=536870912 time=1605.590ms speed=326539.15K/s
> crc16@32: sz=536870912 time=1440.110ms speed=364061.15K/s
> crc16@128: sz=536870912 time=1374.726ms speed=381376.47K/s
:
:
> crc16@536870912: sz=536870912 time=1366.145ms speed=383771.74K/s
This basically doesn't change regardless of the data size, and is only
marginally faster than the optimized crc32c at 4 byte size.
> crc32c-intel@4: sz=536870912 time=1025.192ms speed=511404.47K/s
> crc32c-intel@32: sz=536870912 time=135.124ms speed=3880043.59K/s
> crc32c-intel@128: sz=536870912 time=121.991ms speed=4297766.44K/s
> crc32c-intel@256: sz=536870912 time=120.008ms speed=4368783.42K/s
:
:
> crc32c-intel@536870912: sz=536870912 time=118.369ms speed=4429255.67K/s
This is already faster than crc16 at 4 byte size, and is miles ahead at
32 byte size (the group descriptor CRC chunk size, regardless of whether
a 32-byte or 64-byte descriptor is actually used).
> As you can see from the results, the algorithm(s) that are fast generally
> don't reach full speed until they hit 4KB chunk sizes. powerpc64 and a
> laptop seems to yield speed scaling similar results.
While it is true they don't reach peak speed until 4kB, even at chunk
sizes as small as 32 bytes crc32c is a clear winner. This makes me think
that using CRC32c LSB for the group descriptor checksums when RO_COMPAT_CSUM
is present may be worth the effort.
> crc32c-sby8-* = my new crc32c implementation
> crc16 = kernel's crc16 implementation
> crc32c = kernel's current crc32c sw implementation
> crc16-t10dif = t10dif crc16 implementation
> crc32c-intel = hw accelerated crc32c
>
> --D
Cheers, Andreas