[U-Boot] [PATCH v2 1/7] fs: ext4: do not allow writes if metadata checksum is active

[Jean-Jacques Hiblot]

On 01/02/2019 16:23, Tom Rini wrote:
> On Fri, Feb 01, 2019 at 03:33:34PM +0100, Jean-Jacques Hiblot wrote:
>
>> u-boot does not supports updating the metadata chacksums
>>
>> Signed-off-by: Jean-Jacques Hiblot <jjhiblot at ti.com>
>>
>> ---
>>
>> Changes in v2:
>> - Prevent write access if metadata checksum is enabled
>>
>>   fs/ext4/ext4_write.c | 12 ++++++++++--
>>   include/ext4fs.h     |  1 +
>>   2 files changed, 11 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/ext4/ext4_write.c b/fs/ext4/ext4_write.c
>> index a7f543f7df..1de29236f0 100644
>> --- a/fs/ext4/ext4_write.c
>> +++ b/fs/ext4/ext4_write.c
>> @@ -858,12 +858,19 @@ int ext4fs_write(const char *fname, unsigned char *buffer,
>>   
>>   	g_parent_inode = zalloc(fs->inodesz);
>>   	if (!g_parent_inode)
>> -		goto fail;
>> +		goto fail_ext4fs_init;
>>   
>>   	if (ext4fs_init() != 0) {
>>   		printf("error in File System init\n");
>> -		return -1;
>> +		goto fail_ext4fs_init;
>> +	}
>> +
>> +	if (le32_to_cpu(fs->sb->feature_ro_compat) &
>> +		EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) {
>> +		printf("u-boot doesn't support updating the metadata checksums yet\n");
>> +		goto fail;
>>   	}
>> +
>>   	inodes_per_block = fs->blksz / fs->inodesz;
>>   	parent_inodeno = ext4fs_get_parent_inode_num(fname, filename, F_FILE);
>>   	if (parent_inodeno == -1)
> I'm following up to myself on this one as, Ugh.  To repeat myself from a
> while back:
> commit 6f94ab6656ceffb3f2a972c8de4c554502b6f2b7
> Author: Tom Rini <trini at konsulko.com>
> Date:   Fri Jul 22 17:59:11 2016 -0400
>
>      ext4: Refuse to mount filesystems with 64bit feature set
>      
>      With e2fsprogs after 1.43 the 64bit and metadata_csum features are
>      enabled by default.  The metadata_csum feature changes how
>      ext4_group_desc->bg_checksum is calculated, which would break write
>      support.  The 64bit feature however introduces changes such that it
>      cannot be read by implementations that do not support it.  Since we do
>      not support this, we must not mount it.
>      
>      Cc: Stephen Warren <swarren at nvidia.com>
>      Cc: Simon Glass <sjg at chromium.org>
>      Cc: Lukasz Majewski <l.majewski at samsung.com>
>      Cc: Stefan Roese <sr at denx.de>
>      Reported-by: Andrew Bradford <andrew.bradford at kodakalaris.com>
>      Signed-off-by: Tom Rini <trini at konsulko.com>
>
> Which means that starting way back then I should have also done
> something to say we cannot write to these new images either.  It's good
> and important to finally catch this failure now.  I suspect it's however
> going to start to be an unexpected problem.  Have you any idea how much
> work would go in to supporting the metadata_csum feature?  Thanks again!

I had a look and this is going to be a non-trivial job. The ext4 code is 
showing its age.  And then there would be another couple of flags to 
handle too like journal checksuming, and then also hashtree dirs.

There are a couple of ext4 libs out there (I'm thinking of lwext4 or 
e2fsprogs) that do a much better job at handling the 
compatible/incompatible/read-only flags. I wonder if we wouldn't be 
better off adapting one of them than fixing the current implementation.

JJ

>