Re: how to preallocate??

[Nathan Scott <nathans@xxxxxxx>]

On Wed, Sep 08, 2004 at 01:25:29PM +0200, Miguel Angel de Vega wrote:
> Hi all!
> 
> I'm trying to do a program that preallocate a file and fill it, but when I 
> use 
> xfs_bmap on the file gives me the same number of holes that if I don't do 
> preallocation. To preallocate I use de ioctl call with XFS_IOC_RESVSP64 in 
> this way:

Could you post the "xfs_bmap -v" output after the preallocate?
And would be a good idea to do some error checking in your
program (esp. to make sure the preallocate is succeeding).

> int main(int argc, char **argv)
> {
>       int oflags = O_CREAT|O_TRUNC|O_WRONLY;
>       int fd=0;
>       xfs_flock64_t flck;
>       if(argc < 3)
>       {
>               printf("usage: %s [filename] [numbytes]\n", argv[0]);
>               exit (1);
>       }
>       
>       int numImages = atoll(argv[2]);
>     
>       int aux = 66;
>       long long int bufferSize = 720 * 576 * 2;

Thats an odd write buffer size (829440 bytes?)... why that?

>       int buffer[bufferSize];
>       bzero(buffer, bufferSize);
>       
>       
>       flck.l_whence = SEEK_SET;
>       flck.l_start  = 0LL;
>       flck.l_len    = numImages * bufferSize;
>       
>       fd = open(argv[1], oflags, 0600);
>       int res = ioctl(fd, XFS_IOC_RESVSP64, &flck);

So, add an exit(0) here and then grab the verbose xfs_bmap output,
that'll show the extent layout before your writes.

>       for(int i = 0; i < numImages; i++)
>               write(fd, buffer, bufferSize);
>               
>       close(fd);
>       exit( 0 );
> }


Also, is your filesystem anywhere near full?  Perhaps the
allocator is just unable to find contiguous chunks?

Actually it'll probably be easier to figure out the behavior
you're seeing using xfs_io, e.g:

[root@bruce fsgqa]# xfs_io -f /scratch/xfs5/test
xfs_io> resvsp 0 100m
xfs_io> bmap -v
/scratch/xfs5/test:
 EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET         TOTAL FLAGS
   0: [0..204799]:     104..204903       0 (104..204903)    204800 10000
xfs_io> pwrite -b 829440 0 100m
wrote 104857600/104857600 bytes at offset 0
100 MiB, 127 ops; 00:00:01.375355 (72.709 MiB/sec and 92.3398 ops/sec)
xfs_io> fsync
xfs_io> bmap -v
/scratch/xfs5/test:
 EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET         TOTAL
   0: [0..204799]:     104..204903       0 (104..204903)    204800
xfs_io> 


So, sounds like you're not seeing that sort of layout?
What does the above produce on your filesystem?

cheers.

-- 
Nathan