Re: [PATCH AUTOSEL 5.4 18/55] md/bitmap: don't set sb values if can't pass sanity check

[Sasha Levin <sashal@kernel.org>]

I'm sorry, I couldn't parse the mail below.

On Wed, Jun 01, 2022 at 05:36:15PM -0400, John Stoffel wrote:
>>>>>> "Sasha" == Sasha Levin <sashal@kernel.org> writes:
>
>Sasha> From: Heming Zhao <heming.zhao@suse.com>
>Sasha> [ Upstream commit e68cb83a57a458b01c9739e2ad9cb70b04d1e6d2 ]
>
>Sasha> If bitmap area contains invalid data, kernel will crash then mdadm
>Sasha> triggers "Segmentation fault".
>Sasha> This is cluster-md speical bug. In non-clustered env, mdadm will
>
>special
>
>All the commit messages need to be fixed from what I see.
>
>Sasha> handle broken metadata case. In clustered array, only kernel space
>Sasha> handles bitmap slot info. But even this bug only happened in clustered
>Sasha> env, current sanity check is wrong, the code should be changed.
>
>Sasha> How to trigger: (faulty injection)
>
>Sasha> dd if=/dev/zero bs=1M count=1 oflag=direct of=/dev/sda
>Sasha> dd if=/dev/zero bs=1M count=1 oflag=direct of=/dev/sdb
>Sasha> mdadm -C /dev/md0 -b clustered -e 1.2 -n 2 -l mirror /dev/sda /dev/sdb
>Sasha> mdadm -Ss
>Sasha> echo aaa > magic.txt
>Sasha>  == below modifying slot 2 bitmap data ==
>Sasha> dd if=magic.txt of=/dev/sda seek=16384 bs=1 count=3 <== destroy magic
>Sasha> dd if=/dev/zero of=/dev/sda seek=16436 bs=1 count=4 <== ZERO chunksize
>Sasha> mdadm -A /dev/md0 /dev/sda /dev/sdb
>Sasha>  == kernel crashes. mdadm outputs "Segmentation fault" ==
>
>Sasha> Reason of kernel crash:
>
>Sasha> In md_bitmap_read_sb (called by md_bitmap_create), bad bitmap magic didn't
>Sasha> block chunksize assignment, and zero value made DIV_ROUND_UP_SECTOR_T()
>Sasha> trigger "divide error".
>
>Sasha> Crash log:
>
>Sasha> kernel: md: md0 stopped.
>Sasha> kernel: md/raid1:md0: not clean -- starting background reconstruction
>Sasha> kernel: md/raid1:md0: active with 2 out of 2 mirrors
>Sasha> kernel: dlm: ... ...
>Sasha> kernel: md-cluster: Joined cluster 44810aba-38bb-e6b8-daca-bc97a0b254aa slot 1
>Sasha> kernel: md0: invalid bitmap file superblock: bad magic
>Sasha> kernel: md_bitmap_copy_from_slot can't get bitmap from slot 2
>Sasha> kernel: md-cluster: Could not gather bitmaps from slot 2
>Sasha> kernel: divide error: 0000 [#1] SMP NOPTI
>Sasha> kernel: CPU: 0 PID: 1603 Comm: mdadm Not tainted 5.14.6-1-default
>Sasha> kernel: Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
>Sasha> kernel: RIP: 0010:md_bitmap_create+0x1d1/0x850 [md_mod]
>Sasha> kernel: RSP: 0018:ffffc22ac0843ba0 EFLAGS: 00010246
>Sasha> kernel: ... ...
>Sasha> kernel: Call Trace:
>Sasha> kernel:  ? dlm_lock_sync+0xd0/0xd0 [md_cluster 77fe..7a0]
>Sasha> kernel:  md_bitmap_copy_from_slot+0x2c/0x290 [md_mod 24ea..d3a]
>Sasha> kernel:  load_bitmaps+0xec/0x210 [md_cluster 77fe..7a0]
>Sasha> kernel:  md_bitmap_load+0x81/0x1e0 [md_mod 24ea..d3a]
>Sasha> kernel:  do_md_run+0x30/0x100 [md_mod 24ea..d3a]
>Sasha> kernel:  md_ioctl+0x1290/0x15a0 [md_mod 24ea....d3a]
>Sasha> kernel:  ? mddev_unlock+0xaa/0x130 [md_mod 24ea..d3a]
>Sasha> kernel:  ? blkdev_ioctl+0xb1/0x2b0
>Sasha> kernel:  block_ioctl+0x3b/0x40
>Sasha> kernel:  __x64_sys_ioctl+0x7f/0xb0
>Sasha> kernel:  do_syscall_64+0x59/0x80
>Sasha> kernel:  ? exit_to_user_mode_prepare+0x1ab/0x230
>Sasha> kernel:  ? syscall_exit_to_user_mode+0x18/0x40
>Sasha> kernel:  ? do_syscall_64+0x69/0x80
>Sasha> kernel:  entry_SYSCALL_64_after_hwframe+0x44/0xae
>Sasha> kernel: RIP: 0033:0x7f4a15fa722b
>Sasha> kernel: ... ...
>Sasha> kernel: ---[ end trace 8afa7612f559c868 ]---
>Sasha> kernel: RIP: 0010:md_bitmap_create+0x1d1/0x850 [md_mod]
>
>Sasha> Reported-by: kernel test robot <lkp@intel.com>
>Sasha> Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
>Sasha> Acked-by: Guoqing Jiang <guoqing.jiang@linux.dev>
>Sasha> Signed-off-by: Heming Zhao <heming.zhao@suse.com>
>Sasha> Signed-off-by: Song Liu <song@kernel.org>
>Sasha> Signed-off-by: Sasha Levin <sashal@kernel.org>
>Sasha> ---
>Sasha>  drivers/md/md-bitmap.c | 44 ++++++++++++++++++++++--------------------
>Sasha>  1 file changed, 23 insertions(+), 21 deletions(-)
>
>Sasha> diff --git a/drivers/md/md-bitmap.c b/drivers/md/md-bitmap.c
>Sasha> index d7eef5292ae2..a95e20c3d0d4 100644
>Sasha> --- a/drivers/md/md-bitmap.c
>Sasha> +++ b/drivers/md/md-bitmap.c
>Sasha> @@ -642,14 +642,6 @@ static int md_bitmap_read_sb(struct bitmap *bitmap)
>Sasha>  	daemon_sleep = le32_to_cpu(sb->daemon_sleep) * HZ;
>Sasha>  	write_behind = le32_to_cpu(sb->write_behind);
>Sasha>  	sectors_reserved = le32_to_cpu(sb->sectors_reserved);
>Sasha> -	/* Setup nodes/clustername only if bitmap version is
>Sasha> -	 * cluster-compatible
>Sasha> -	 */
>Sasha> -	if (sb->version == cpu_to_le32(BITMAP_MAJOR_CLUSTERED)) {
>Sasha> -		nodes = le32_to_cpu(sb->nodes);
>Sasha> -		strlcpy(bitmap->mddev->bitmap_info.cluster_name,
>Sasha> -				sb->cluster_name, 64);
>Sasha> -	}
>
>Sasha>  	/* verify that the bitmap-specific fields are valid */
>Sasha>  	if (sb->magic != cpu_to_le32(BITMAP_MAGIC))
>Sasha> @@ -671,6 +663,16 @@ static int md_bitmap_read_sb(struct bitmap *bitmap)
>Sasha>  		goto out;
>Sasha>  	}
>
>Sasha> +	/*
>Sasha> +	 * Setup nodes/clustername only if bitmap version is
>Sasha> +	 * cluster-compatible
>Sasha> +	 */
>Sasha> +	if (sb->version == cpu_to_le32(BITMAP_MAJOR_CLUSTERED)) {
>Sasha> +		nodes = le32_to_cpu(sb->nodes);
>Sasha> +		strlcpy(bitmap->mddev->bitmap_info.cluster_name,
>Sasha> +				sb->cluster_name, 64);
>Sasha> +	}
>Sasha> +
>Sasha>  	/* keep the array size field of the bitmap superblock up to date */
>sb-> sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);
>
>Sasha> @@ -703,9 +705,9 @@ static int md_bitmap_read_sb(struct bitmap *bitmap)
>
>Sasha>  out:
>Sasha>  	kunmap_atomic(sb);
>Sasha> -	/* Assigning chunksize is required for "re_read" */
>Sasha> -	bitmap->mddev->bitmap_info.chunksize = chunksize;
>Sasha>  	if (err == 0 && nodes && (bitmap->cluster_slot < 0)) {
>Sasha> +		/* Assigning chunksize is required for "re_read" */
>Sasha> +		bitmap->mddev->bitmap_info.chunksize = chunksize;
>Sasha>  		err = md_setup_cluster(bitmap->mddev, nodes);
>Sasha>  		if (err) {
>Sasha>  			pr_warn("%s: Could not setup cluster service (%d)\n",
>Sasha> @@ -716,18 +718,18 @@ static int md_bitmap_read_sb(struct bitmap *bitmap)
>Sasha>  		goto re_read;
>Sasha>  	}
>
>Sasha> -
>Sasha>  out_no_sb:
>Sasha> -	if (test_bit(BITMAP_STALE, &bitmap->flags))
>Sasha> -		bitmap->events_cleared = bitmap->mddev->events;
>Sasha> -	bitmap->mddev->bitmap_info.chunksize = chunksize;
>Sasha> -	bitmap->mddev->bitmap_info.daemon_sleep = daemon_sleep;
>Sasha> -	bitmap->mddev->bitmap_info.max_write_behind = write_behind;
>Sasha> -	bitmap->mddev->bitmap_info.nodes = nodes;
>Sasha> -	if (bitmap->mddev->bitmap_info.space == 0 ||
>Sasha> -	    bitmap->mddev->bitmap_info.space > sectors_reserved)
>Sasha> -		bitmap->mddev->bitmap_info.space = sectors_reserved;
>Sasha> -	if (err) {
>Sasha> +	if (err == 0) {
>Sasha> +		if (test_bit(BITMAP_STALE, &bitmap->flags))
>Sasha> +			bitmap->events_cleared = bitmap->mddev->events;
>Sasha> +		bitmap->mddev->bitmap_info.chunksize = chunksize;
>Sasha> +		bitmap->mddev->bitmap_info.daemon_sleep = daemon_sleep;
>Sasha> +		bitmap->mddev->bitmap_info.max_write_behind = write_behind;
>Sasha> +		bitmap->mddev->bitmap_info.nodes = nodes;
>Sasha> +		if (bitmap->mddev->bitmap_info.space == 0 ||
>Sasha> +			bitmap->mddev->bitmap_info.space > sectors_reserved)
>Sasha> +			bitmap->mddev->bitmap_info.space = sectors_reserved;
>Sasha> +	} else {
>Sasha>  		md_bitmap_print_sb(bitmap);
>Sasha>  		if (bitmap->cluster_slot < 0)
>Sasha>  			md_cluster_stop(bitmap->mddev);
>Sasha> --
>Sasha> 2.35.1
>

-- 
Thanks,
Sasha