Re: [PATCH] resize2fs vs. large inodes, take 2

[Eric Sandeen <sandeen@redhat.com>]

Andreas Dilger wrote:

>> ===================================================================
>> --- e2fsprogs-1.40.7.orig/resize/resize2fs.c
>> +++ e2fsprogs-1.40.7/resize/resize2fs.c
>> @@ -1168,11 +1168,12 @@ static errcode_t inode_scan_and_fix(ext2
>>  	 * elsewhere in the inode table
>>  	 */
>>  	while (1) {
>> -		retval = ext2fs_get_next_inode(scan, &ino, &inode);
>> +		retval = ext2fs_get_next_inode_full(scan, &ino, buf, inode_size);
>>  		if (retval) goto errout;
>>  		if (!ino)
>>  			break;
>>  
>> +		memcpy(&inode, buf, sizeof(struct ext2_inode));
> 
> Should this be using "sizeof(struct ext2_inode)" or should it be using
> "sb->s_inode_size" instead (extracted from the right struct of course)?

well, let's see... I think we read "inode_size" in get_next_inode_full,
which is s_inode_size, into buf, which was allocated to size
inode_size/s_inode_size.

But "inode" is just a plain ol' little inode.  I think really this
"inode" is just for convenience for accessing the normal inode fields....

But I now that I try livecd-creator with this patch, even on 128-byte
inodes, the fscks it runs is finding trouble post-resize (this despite
all the regression test passing...)  *sigh* I think I'd better sit on
this problem for a while longer before I send the next patch :)

Ted, pls ignore this for now...

-Eric