Re: fs/ocfs2/suballoc.c:2430:2-8: preceding lock on line 2413

[Thomas Gleixner]

On Fri, Aug 14 2020 at 14:11, kernel test robot wrote:
>
> coccinelle warnings: (new ones prefixed by >>)
>
>>> fs/ocfs2/suballoc.c:2430:2-8: preceding lock on line 2413
>
> vim +2430 fs/ocfs2/suballoc.c
>
> 415cb800375cc4 Mark Fasheh     2007-09-16  2379  
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2380  static int ocfs2_block_group_clear_bits(handle_t *handle,
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2381  					struct inode *alloc_inode,
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2382  					struct ocfs2_group_desc *bg,
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2383  					struct buffer_head *group_bh,
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2384  					unsigned int bit_off,
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2385  					unsigned int num_bits,
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2386  					void (*undo_fn)(unsigned int bit,
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2387  							unsigned long *bmap))
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2388  {
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2389  	int status;
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2390  	unsigned int tmp;
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2391  	struct ocfs2_group_desc *undo_bg = NULL;
> 464170647b5648 Thomas Gleixner 2019-08-09  2392  	struct journal_head *jh;
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2393  
> 42035306132802 Joel Becker     2008-11-13  2394  	/* The caller got this descriptor from
> 42035306132802 Joel Becker     2008-11-13  2395  	 * ocfs2_read_group_descriptor().  Any corruption is a code bug. */
> 42035306132802 Joel Becker     2008-11-13  2396  	BUG_ON(!OCFS2_IS_VALID_GROUP_DESC(bg));
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2397  
> 2f73e135b83c50 Tao Ma          2011-02-22  2398  	trace_ocfs2_block_group_clear_bits(bit_off, num_bits);
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2399  
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2400  	BUG_ON(undo_fn && !ocfs2_is_cluster_bitmap(alloc_inode));
> 0cf2f7632b1789 Joel Becker     2009-02-12  2401  	status = ocfs2_journal_access_gd(handle, INODE_CACHE(alloc_inode),
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2402  					 group_bh,
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2403  					 undo_fn ?
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2404  					 OCFS2_JOURNAL_ACCESS_UNDO :
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2405  					 OCFS2_JOURNAL_ACCESS_WRITE);
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2406  	if (status < 0) {
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2407  		mlog_errno(status);
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2408  		goto bail;
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2409  	}
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2410  
> 464170647b5648 Thomas Gleixner 2019-08-09  2411  	jh = bh2jh(group_bh);
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2412  	if (undo_fn) {
> 464170647b5648 Thomas Gleixner 2019-08-09 @2413  		spin_lock(&jh->b_state_lock);
> 464170647b5648 Thomas Gleixner 2019-08-09  2414  		undo_bg = (struct ocfs2_group_desc *) jh->b_committed_data;
> 94e41ecfe0f202 Sunil Mushran   2009-06-19  2415  		BUG_ON(!undo_bg);
> 94e41ecfe0f202 Sunil Mushran   2009-06-19  2416  	}
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2417  
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2418  	tmp = num_bits;
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2419  	while(tmp--) {
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2420  		ocfs2_clear_bit((bit_off + tmp),
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2421  				(unsigned long *) bg->bg_bitmap);
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2422  		if (undo_fn)
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2423  			undo_fn(bit_off + tmp,
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2424  				(unsigned long *) undo_bg->bg_bitmap);
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2425  	}
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2426  	le16_add_cpu(&bg->bg_free_bits_count, num_bits);
> 9b5cd10e4c14a1 Srinivas Eeda   2010-10-05  2427  	if (le16_to_cpu(bg->bg_free_bits_count) > le16_to_cpu(bg->bg_bits)) {
> e75ed71be4f2f7 Changwei Ge     2018-01-31  2428  		if (undo_fn)
> 464170647b5648 Thomas Gleixner 2019-08-09  2429  			spin_unlock(&jh->b_state_lock);
> 7ecef14ab1db96 Joe Perches     2015-09-04 @2430  		return ocfs2_error(alloc_inode->i_sb, "Group descriptor # %llu has bit count %u but claims %u are freed. num_bits %d\n",
> 9b5cd10e4c14a1 Srinivas Eeda   2010-10-05  2431  				   (unsigned long long)le64_to_cpu(bg->bg_blkno),
> 9b5cd10e4c14a1 Srinivas Eeda   2010-10-05  2432  				   le16_to_cpu(bg->bg_bits),
> 7ecef14ab1db96 Joe Perches     2015-09-04  2433  				   le16_to_cpu(bg->bg_free_bits_count),
> 7ecef14ab1db96 Joe Perches     2015-09-04  2434  				   num_bits);
> 9b5cd10e4c14a1 Srinivas Eeda   2010-10-05  2435  	}
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2436  
> b4414eea0e7b9c Mark Fasheh     2010-03-11  2437  	if (undo_fn)
> 464170647b5648 Thomas Gleixner 2019-08-09  2438  		spin_unlock(&jh->b_state_lock);
> 94e41ecfe0f202 Sunil Mushran   2009-06-19  2439  
> ec20cec7a35158 Joel Becker     2010-03-19  2440  	ocfs2_journal_dirty(handle, group_bh);
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2441  bail:
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2442  	return status;
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2443  }
> ccd979bdbce9fb Mark Fasheh     2005-12-15  2444  

That's clearly a false positive. Is there anything what can be done to
help that cocci script here?

Thanks,

        tglx