Re: [PATCH] erofs: code for verifying superblock checksum of an erofs image.

From: Gao Xiang
Date: Mon Oct 21 2019 - 02:22:35 EST


Hi Pratik,

Some comments as below...

On Mon, Oct 21, 2019 at 12:58:28AM +0530, Pratik Shinde wrote:
> Patch for kernel side changes of checksum feature.I used kernel's
> crc32c library for calculating the checksum.
>
> Signed-off-by: Pratik Shinde <pratikshinde320@xxxxxxxxx>
> ---
> fs/erofs/erofs_fs.h | 5 +++--
> fs/erofs/internal.h | 2 +-
> fs/erofs/super.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++----
> 3 files changed, 50 insertions(+), 7 deletions(-)
>
> diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
> index b1ee565..bab5506 100644
> --- a/fs/erofs/erofs_fs.h
> +++ b/fs/erofs/erofs_fs.h
> @@ -17,6 +17,7 @@
> */
> #define EROFS_FEATURE_INCOMPAT_LZ4_0PADDING 0x00000001
> #define EROFS_ALL_FEATURE_INCOMPAT EROFS_FEATURE_INCOMPAT_LZ4_0PADDING
> +#define EROFS_FEATURE_SB_CHKSUM 0x0001

How about keeping in sync with erofs-utils?

>
> /* 128-byte erofs on-disk super block */
> struct erofs_super_block {
> @@ -37,8 +38,8 @@ struct erofs_super_block {
> __u8 uuid[16]; /* 128-bit uuid for volume */
> __u8 volume_name[16]; /* volume name */
> __le32 feature_incompat;
> -
> - __u8 reserved2[44];
> + __le32 chksum_blocks; /* number of blocks used for checksum */
> + __u8 reserved2[40];
> };
>
> /*
> diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
> index 544a453..cd3af45 100644
> --- a/fs/erofs/internal.h
> +++ b/fs/erofs/internal.h
> @@ -86,7 +86,7 @@ struct erofs_sb_info {
> u8 uuid[16]; /* 128-bit uuid for volume */
> u8 volume_name[16]; /* volume name */
> u32 feature_incompat;
> -
> + u32 features;

no use? and
how about keeping in sync with erofs-utils, naming feature_compat as well? see,

https://lore.kernel.org/r/20191020141102.GA30399@hsiangkao-HP-ZHAN-66-Pro-G1/

> unsigned int mount_opt;
> };
>
> diff --git a/fs/erofs/super.c b/fs/erofs/super.c
> index 0e36949..94e1d6a 100644
> --- a/fs/erofs/super.c
> +++ b/fs/erofs/super.c
> @@ -9,6 +9,7 @@
> #include <linux/statfs.h>
> #include <linux/parser.h>
> #include <linux/seq_file.h>
> +#include <linux/crc32c.h>
> #include "xattr.h"
>
> #define CREATE_TRACE_POINTS
> @@ -46,6 +47,45 @@ void _erofs_info(struct super_block *sb, const char *function,
> va_end(args);
> }
>
> +static int erofs_validate_sb_chksum(struct erofs_super_block *dsb,
> + struct super_block *sb)

erofs_validate_sb_chksum(struct super_block *sb, void *sbdata)

Can we pass void *sbdata --- first page(block) instead?

> +{
> + u32 disk_chksum = le32_to_cpu(dsb->checksum);
> + u32 nblocks = le32_to_cpu(dsb->chksum_blocks);
> + u32 crc;
> + struct erofs_super_block *dsb2;
> + char *buf;
> + unsigned int off = 0;
> + void *kaddr;
> + struct page *page;
> + int i, ret = -EINVAL;
> +
> + buf = kmalloc(nblocks * EROFS_BLKSIZ, GFP_KERNEL);

IMO, I think we could just
1) malloc EROFS_BLKSIZ;
2) copy sbdata -> buf;
3) dsb->checksum = 0;
4) crc32c the first copied block;
5) kfree.

or some better approach without allocating memory.

> + if (!buf)
> + goto out;

how about erroring out with proper return value?


> + for (i = 0; i < nblocks; i++) {
> + page = erofs_get_meta_page(sb, i);
> + if (IS_ERR(page))

ditto.

> + goto out;
> + kaddr = kmap_atomic(page);
> + (void) memcpy(buf + off, kaddr, EROFS_BLKSIZ);

we can call crc32c for multiple times, and no need to
get and calc 0-th block again...

and no memcpy at all.

> + kunmap_atomic(kaddr);
> + unlock_page(page);
> + /* first page will be released by erofs_read_superblock */
> + if (i != 0)
> + put_page(page);
> + off += EROFS_BLKSIZ;
> + }
> + dsb2 = (struct erofs_super_block *)(buf + EROFS_SUPER_OFFSET);
> + dsb2->checksum = 0;
> + crc = crc32c(0, buf, nblocks * EROFS_BLKSIZ);
> + if (crc != disk_chksum)
> + goto out;
> + ret = 0;
> +out: kfree(buf);
> + return ret;
> +}
> +
> static void erofs_inode_init_once(void *ptr)
> {
> struct erofs_inode *vi = ptr;
> @@ -109,18 +149,20 @@ static int erofs_read_superblock(struct super_block *sb)
> erofs_err(sb, "cannot read erofs superblock");
> return PTR_ERR(page);
> }
> -
> sbi = EROFS_SB(sb);
> -
> data = kmap_atomic(page);

should use kmap() then, because kmalloc() and get_meta_page() can sleep and
last for long time.

> dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
> -
> ret = -EINVAL;
> if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
> erofs_err(sb, "cannot find valid erofs superblock");
> goto out;
> }
> -
> + if (dsb->feature_compat & EROFS_FEATURE_SB_CHKSUM) {
> + if (erofs_validate_sb_chksum(dsb, sb)) {

erofs_validate_sb_chksum(sb, data)

> + erofs_err(sb, "super block checksum incorrect");

It should be with a proper return value and corresponding messages,
see the related discussion:
https://lore.kernel.org/r/20190822090541.GA193349@architecture4/

Thanks,
Gao Xiang

> + goto out;
> + }
> + }
> blkszbits = dsb->blkszbits;
> /* 9(512 bytes) + LOG_SECTORS_PER_BLOCK == LOG_BLOCK_SIZE */
> if (blkszbits != LOG_BLOCK_SIZE) {
> --
> 2.9.3
>