linux-kernel - Re: [git pull] first batch of ufs fixes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 14 Jun 2017 13:33:29 -0700 (PDT)
From:   Richard Narron <comet.berkeley@...il.com>
To:     Al Viro <viro@...IV.linux.org.uk>
cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org
Subject: Re: [git pull] first batch of ufs fixes

On Wed, 14 Jun 2017, Al Viro wrote:
...
> AFAICS, a conservative approach would be
> 	* reject UFS_42POSTBLFMT for 44bsd ones - it's almost certainly
> *not* one.
> 	* check if fs_maxbsize is equal to frag size; treat that as
> "counts are read from new location and stored both to old and new".
> 44bsd fs_maxbsize != block size => not converted, just use old locations
> for everything.  UFS2 => use new locations for everything, don't bother
> with old ones.  IOW, something like this (WARNING: completely untested,
> might screw your filesystem) might do.
>
> NOTE: all I have is your image *after* it had counters buggered; I don't
> know the exact sequence of operations that fucked it in your case.  One
> way to trigger it is to mount/umount on OpenBSD, then mount/modify/umount
> on Linux, then mount/umount on OpenBSD, then fsck on OpenBSD.  This patch
> apparently fixes that, but your reproducer might be something different.
>
> Signed-off-by: Al Viro <viro@...iv.linux.org.uk>
> ---
> diff --git a/fs/ufs/super.c b/fs/ufs/super.c
> index d9aa2627c9df..eca838a8b43e 100644
> --- a/fs/ufs/super.c
> +++ b/fs/ufs/super.c
> @@ -480,7 +480,7 @@ static void ufs_setup_cstotal(struct super_block *sb)
> 	usb3 = ubh_get_usb_third(uspi);
>
> 	if ((mtype == UFS_MOUNT_UFSTYPE_44BSD &&
> -	     (usb1->fs_flags & UFS_FLAGS_UPDATED)) ||
> +	     (usb2->fs_un.fs_u2.fs_maxbsize == usb1->fs_bsize)) ||
> 	    mtype == UFS_MOUNT_UFSTYPE_UFS2) {
> 		/*we have statistic in different place, then usual*/
> 		uspi->cs_total.cs_ndir = fs64_to_cpu(sb, usb2->fs_un.fs_u2.cs_ndir);
> @@ -596,9 +596,7 @@ static void ufs_put_cstotal(struct super_block *sb)
> 	usb2 = ubh_get_usb_second(uspi);
> 	usb3 = ubh_get_usb_third(uspi);
>
> -	if ((mtype == UFS_MOUNT_UFSTYPE_44BSD &&
> -	     (usb1->fs_flags & UFS_FLAGS_UPDATED)) ||
> -	    mtype == UFS_MOUNT_UFSTYPE_UFS2) {
> +	if (mtype == UFS_MOUNT_UFSTYPE_UFS2) {
> 		/*we have statistic in different place, then usual*/
> 		usb2->fs_un.fs_u2.cs_ndir =
> 			cpu_to_fs64(sb, uspi->cs_total.cs_ndir);
> @@ -608,16 +606,26 @@ static void ufs_put_cstotal(struct super_block *sb)
> 			cpu_to_fs64(sb, uspi->cs_total.cs_nifree);
> 		usb3->fs_un1.fs_u2.cs_nffree =
> 			cpu_to_fs64(sb, uspi->cs_total.cs_nffree);
> -	} else {
> -		usb1->fs_cstotal.cs_ndir =
> -			cpu_to_fs32(sb, uspi->cs_total.cs_ndir);
> -		usb1->fs_cstotal.cs_nbfree =
> -			cpu_to_fs32(sb, uspi->cs_total.cs_nbfree);
> -		usb1->fs_cstotal.cs_nifree =
> -			cpu_to_fs32(sb, uspi->cs_total.cs_nifree);
> -		usb1->fs_cstotal.cs_nffree =
> -			cpu_to_fs32(sb, uspi->cs_total.cs_nffree);
> +		goto out;
> 	}
> +
> +	if (mtype == UFS_MOUNT_UFSTYPE_44BSD &&
> +	     (usb2->fs_un.fs_u2.fs_maxbsize == usb1->fs_bsize)) {
> +		/* store stats in both old and new places */
> +		usb2->fs_un.fs_u2.cs_ndir =
> +			cpu_to_fs64(sb, uspi->cs_total.cs_ndir);
> +		usb2->fs_un.fs_u2.cs_nbfree =
> +			cpu_to_fs64(sb, uspi->cs_total.cs_nbfree);
> +		usb3->fs_un1.fs_u2.cs_nifree =
> +			cpu_to_fs64(sb, uspi->cs_total.cs_nifree);
> +		usb3->fs_un1.fs_u2.cs_nffree =
> +			cpu_to_fs64(sb, uspi->cs_total.cs_nffree);
> +	}
> +	usb1->fs_cstotal.cs_ndir = cpu_to_fs32(sb, uspi->cs_total.cs_ndir);
> +	usb1->fs_cstotal.cs_nbfree = cpu_to_fs32(sb, uspi->cs_total.cs_nbfree);
> +	usb1->fs_cstotal.cs_nifree = cpu_to_fs32(sb, uspi->cs_total.cs_nifree);
> +	usb1->fs_cstotal.cs_nffree = cpu_to_fs32(sb, uspi->cs_total.cs_nffree);
> +out:
> 	ubh_mark_buffer_dirty(USPI_UBH(uspi));
> 	ufs_print_super_stuff(sb, usb1, usb2, usb3);
> 	UFSD("EXIT\n");
> @@ -997,6 +1005,13 @@ static int ufs_fill_super(struct super_block *sb, void *data, int silent)
> 		flags |=  UFS_ST_SUN;
> 	}
>
> +	if ((flags & UFS_ST_MASK) == UFS_ST_44BSD &&
> +	    uspi->s_postblformat == UFS_42POSTBLFMT) {
> +		if (!silent)
> +			pr_err("this is not a 44bsd filesystem");
> +		goto failed;
> +	}
> +
> 	/*
> 	 * Check ufs magic number
> 	 */
>

Al this patch looks good to me (so far). I tested all 6 combinations of 
ufs1 and ufs2 in FreeBSD 11.0, OpenBSD 6.1 and NetBSD 7.1

For each combination, I do 5 steps:

   1) BSD: Make a ufs filesystem
      dd if=/dev/zero to the BSD subpartition
      make (newfs) a ufs (1 or 2) filesystem on the BSD subpartiton

   2) Linux: Create a subdirectory and make a large file
      mkdir a
      dd if=/dev/zero bs=1M count=3072

   3) BSD: Check a ufs filesystem
      fsck -f

   4) Linux: Remove the large file and the subdirectory
      rm
      rmdir

   5) BSD; check a ufs filesystem
      fsck -f

Tested-By: Richard Narron <comet.berkeley@...il.com>