Re: [Ocfs2-devel] [PATCH v3 4/4] ocfs2: check/fix inode block for online file check

[Gang He]

Hello Mark,


>>> 
> On Fri, Dec 25, 2015 at 03:16:19PM +0800, Gang He wrote:
>> Implement online check or fix inode block during
>> reading a inode block to memory.
>> 
>> Signed-off-by: Gang He <ghe@xxxxxxxx>
>> ---
>>  fs/ocfs2/inode.c       | 200 
> +++++++++++++++++++++++++++++++++++++++++++++++--
>>  fs/ocfs2/ocfs2_trace.h |   2 +
>>  2 files changed, 196 insertions(+), 6 deletions(-)
>> 
>> diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
>> index 8f87e05..6ac2f19 100644
>> --- a/fs/ocfs2/inode.c
>> +++ b/fs/ocfs2/inode.c
>> @@ -53,6 +53,7 @@
>>  #include "xattr.h"
>>  #include "refcounttree.h"
>>  #include "ocfs2_trace.h"
>> +#include "filecheck.h"
>>  
>>  #include "buffer_head_io.h"
>>  
>> @@ -74,6 +75,14 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super 
> *osb,
>>  				    struct inode *inode,
>>  				    struct buffer_head *fe_bh);
>>  
>> +static int ocfs2_filecheck_read_inode_block_full(struct inode *inode,
>> +						 struct buffer_head **bh,
>> +						 int flags, int type);
>> +static int ocfs2_filecheck_validate_inode_block(struct super_block *sb,
>> +						struct buffer_head *bh);
>> +static int ocfs2_filecheck_repair_inode_block(struct super_block *sb,
>> +					      struct buffer_head *bh);
>> +
>>  void ocfs2_set_inode_flags(struct inode *inode)
>>  {
>>  	unsigned int flags = OCFS2_I(inode)->ip_attr;
>> @@ -127,6 +136,7 @@ struct inode *ocfs2_ilookup(struct super_block *sb, u64 
> blkno)
>>  struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned 
> flags,
>>  			 int sysfile_type)
>>  {
>> +	int rc = 0;
>>  	struct inode *inode = NULL;
>>  	struct super_block *sb = osb->sb;
>>  	struct ocfs2_find_inode_args args;
>> @@ -161,12 +171,17 @@ struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 
> blkno, unsigned flags,
>>  	}
>>  	trace_ocfs2_iget5_locked(inode->i_state);
>>  	if (inode->i_state & I_NEW) {
>> -		ocfs2_read_locked_inode(inode, &args);
>> +		rc = ocfs2_read_locked_inode(inode, &args);
>>  		unlock_new_inode(inode);
>>  	}
>>  	if (is_bad_inode(inode)) {
>>  		iput(inode);
>> -		inode = ERR_PTR(-ESTALE);
>> +		if ((flags & OCFS2_FI_FLAG_FILECHECK_CHK) ||
>> +		    (flags & OCFS2_FI_FLAG_FILECHECK_FIX))
>> +			/* Return OCFS2_FILECHECK_ERR_XXX related errno */
>> +			inode = ERR_PTR(rc);
>> +		else
>> +			inode = ERR_PTR(-ESTALE);
>>  		goto bail;
>>  	}
>>  
>> @@ -494,16 +509,32 @@ static int ocfs2_read_locked_inode(struct inode *inode,
>>  	}
>>  
>>  	if (can_lock) {
>> -		status = ocfs2_read_inode_block_full(inode, &bh,
>> -						     OCFS2_BH_IGNORE_CACHE);
>> +		if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_CHK)
>> +			status = ocfs2_filecheck_read_inode_block_full(inode,
>> +						&bh, OCFS2_BH_IGNORE_CACHE, 0);
>> +		else if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_FIX)
>> +			status = ocfs2_filecheck_read_inode_block_full(inode,
>> +						&bh, OCFS2_BH_IGNORE_CACHE, 1);
>> +		else
>> +			status = ocfs2_read_inode_block_full(inode,
>> +						&bh, OCFS2_BH_IGNORE_CACHE);
>>  	} else {
>>  		status = ocfs2_read_blocks_sync(osb, args->fi_blkno, 1, &bh);
>>  		/*
>>  		 * If buffer is in jbd, then its checksum may not have been
>>  		 * computed as yet.
>>  		 */
>> -		if (!status && !buffer_jbd(bh))
>> -			status = ocfs2_validate_inode_block(osb->sb, bh);
>> +		if (!status && !buffer_jbd(bh)) {
>> +			if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_CHK)
>> +				status = ocfs2_filecheck_validate_inode_block(
>> +								osb->sb, bh);
>> +			else if (args->fi_flags & OCFS2_FI_FLAG_FILECHECK_FIX)
>> +				status = ocfs2_filecheck_repair_inode_block(
>> +								osb->sb, bh);
>> +			else
>> +				status = ocfs2_validate_inode_block(
>> +								osb->sb, bh);
>> +		}
>>  	}
>>  	if (status < 0) {
>>  		mlog_errno(status);
>> @@ -531,6 +562,14 @@ static int ocfs2_read_locked_inode(struct inode *inode,
>>  
>>  	BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));
>>  
>> +	if (buffer_dirty(bh)) {
>> +		status = ocfs2_write_block(osb, bh, INODE_CACHE(inode));
>> +		if (status < 0) {
>> +			mlog_errno(status);
>> +			goto bail;
>> +		}
>> +	}
> 
> This reminds me, we should be checking for a readonly file system up top in
> the 'fix' helper.
When the file system becomes read-only, we still can check certain file via it's inode number and report the result (the inode block is integrated or not),
but we cannot fix this file in file system read-only status, otherwise the subsequent logic will become too complex in case we modify/write a inode block under a read-only file system.
Actually, online file check feature should be combined with "errors=continues" option for better using.

> 
> Also, I'm concerned that the buffer in question might be journaled. In that 
> case, writing it to disk like this could cause corruptions (if the buffer
> contains not-committed changes). 
I ever though of journaling this changed inode block in case file check fixing, but you know, we are being on inode block loading stage, the journal related structs are not prepared at this moment, then I write this block back to the disk synchronously within ocfs2_inode_lock, it looks a little tricky, but not bring any risk. in case the machine crashes when writing the inode block back to the disk, this will not affect file system integrity, since this inode block original is corrupted, the user can fix this inode block via file check again after the machine is recovered.
Anyway, I just want to let all know what I think behind this part code, maybe it is not right, please give your feedback again.

Thanks
Gang

 
> 
> The answer might be to journal the changes we make but I'm not sure if that
> can deadlock with other iget() calls.
> 	--Mark
> 
> --
> Mark Fasheh