Re: [PATCH] fstests: btrfs/002: fix the OOM caused by too large block size

[Qu Wenruo <wqu@suse.com>]

在 2024/10/22 21:54, Anand Jain 写道:
> 
> looks good.
> Nits:
> 
>> For 64K sector size, the BLKS is 512, and BLKS is 128 (one 64K sector).
>                                              ^^ NBLKS
> 
>> $FALLOC is the correct value of 64K (10K rounded up to 64K).
>>
> 
> Not sure what you meant here. More below.

I just mean the old $FALLOC value is 64K, which is correct.

I didn't see anything confusing here, I was explaining each variable 
involved.

> 
> 
>> diff --git a/tests/btrfs/002 b/tests/btrfs/002
>> index f223cc60..0c231b89 100755
>> --- a/tests/btrfs/002
>> +++ b/tests/btrfs/002
>> @@ -70,19 +70,14 @@ _read_modify_write()
>>   _fill_blk()
>>   {
>>       local FSIZE
>> -    local BLKS
>> -    local NBLK
>> -    local FALLOC
>> -    local WS
>> +    local NEWSIZE
>>       for i in `find /$1 -type f`
>>       do
>>           FSIZE=`stat -t $i | cut -d" " -f2`
>> -        BLKS=`stat -c "%B" $i`
>> -        NBLK=`stat -c "%b" $i`
>> -        FALLOC=$(($BLKS * $NBLK))
>> -        WS=$(($FALLOC - $FSIZE))
>> -        _ddt of=$i oseek=$FSIZE obs=$WS count=1 status=noxfer 2>/dev/ 
>> null &
> 
> 
>> +        NEWSIZE=$(( ($FSIZE + $blksize -1 ) / $blksize * $blksize ))
> 
> 
> Please add extra ( ) otherwise it is very confusing.
> 
>     NEWSIZE=$(( (($FSIZE + $blksize -1 ) / $blksize) * $blksize ))

I didn't see how it's confusing.

It's the very basic round down, and I didn't see any kernel code doing 
extra () just for round down.

Thanks,
Qu
> 
> 
> 
>> +
>> +        $XFS_IO_PROG -c "pwrite -i /dev/urandom $FSIZE $(( $NEWSIZE - 
>> $FSIZE ))" $i > /dev/null &
> 
> 
> 
> Reviewed-by: Anand Jain <anand.jain@oracle.com>
> 
> 
> Thanks, Anand