linux-ext4 - Re: [PATCH] ext4: fix quota accounting in case of fallocate

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <87eij1ccvk.fsf@openvz.org>
Date:	Wed, 31 Mar 2010 08:51:27 +0400
From:	dmonakhov@...nvz.org
To:	"Aneesh Kumar K. V" <aneesh.kumar@...ux.vnet.ibm.com>
Cc:	linux-ext4@...r.kernel.org, tytso@....edu
Subject: Re: [PATCH] ext4: fix quota accounting in case of fallocate

"Aneesh Kumar K. V" <aneesh.kumar@...ux.vnet.ibm.com> writes:

> On Tue, 30 Mar 2010 18:24:35 +0400, Dmitry Monakhov <dmonakhov@...nvz.org> wrote:
>> allocated_meta_data is already included in 'used' variable.
>> 
>> Signed-off-by: Dmitry Monakhov <dmonakhov@...nvz.org>
>> ---
>>  fs/ext4/inode.c |    3 ++-
>>  1 files changed, 2 insertions(+), 1 deletions(-)
>> 
>> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
>> index bec222c..bf989fb 100644
>> --- a/fs/ext4/inode.c
>> +++ b/fs/ext4/inode.c
>> @@ -1110,7 +1110,8 @@ void ext4_da_update_reserve_space(struct inode *inode,
>>  		 */
>>  		if (allocated_meta_blocks)
>>  			dquot_claim_block(inode, allocated_meta_blocks);
>> -		dquot_release_reservation_block(inode, mdb_free + used);
>> +		dquot_release_reservation_block(inode, mdb_free + used -
>> +			allocated_meta_blocks);
>>  	}
>> 
>>  	/*
>
> Do we really need to do this ? IIUC reservation count and actual
> allocated count are two different. One block allocation we need to
> remove all the blocks reserved from the reservation count and add
Yes. remove all, but minus already allocated_metadata, which was
accounted in to metadata reservation.
> actually allocated blocks to the allocated count.
Just try an example:
reserve_space (inode, lblock := 1024 ) {
   md_needed = ext4_calc_metadata_amount(inode, lblock) (let it be '2')
   dquot_reserve_block(inode, md_needed + 1) /* '3' i.e blocks  reserved*/
   /* If this is first reservation for this inode then
    dq_rsv = inode->i_reserved_data_blocks + inode->i_reserved_meta_block
   */
}
Later called from fallocate
update_rerved_space(inode, used:=1, claim :=0) {
  /* Let i_allocatd_meta_data is '1' (as it so in most cases)  */
  ei->i_reserved_data_blocks -= used;  /* 1 - 1 => 0 */
  used += ei->i_allocated_meta_blocks; /* 1 + 1 => 2 */
  ei->i_reserved_meta_blocks -= ei->i_allocated_meta_blocks /* 2 - 1 => 1 */
  allocated_meta_blocks = ei->i_allocated_meta_blocks; /*  1 */
  ei->i_allocated_meta_blocks = 0;

  if (ei->i_reserved_data_blocks == 0) /* True in our case */
     mdb_free = ei->i_reserved_meta_blocks; /* mbd_free == 1*/

  if (allocated_meta_blocks)
         dquot_claim_block(inode, allocated_meta_blocks); /* claim '1' block*/
  dquot_release_reservation_block(inode, mdb_free + used); /* free (1 + 2) */

  /* So we reserved:
     dq_rsv = i_reserved_data_blocks + i_reserved_meta_block ( 3 blocks)
     But during update we claim + free:
          i_allocated_meta_data+(i_reserved_data_block+i_reserved_meta_data)
          (4 blocks).
          Which result in incorrect dquota reservation accounting(it
          goes negative)
  */

Initially i've found the issue by executing fsstress with falloc support.
It takes enouth process to catch writepage/fallocate overlapping.
xfstests-dev/ltp/xfsfsstress -p100 -n99999999

--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html