[pnfs] [PATCH 04/28] pnfs: FIXME: NFS4_PNFS_DEV_MAXSIZE too small

Fredric Isaman iisaman at citi.umich.edu
Fri Mar 14 08:58:28 EDT 2008



On Thu, 13 Mar 2008, Dean Hildebrand wrote:

>
>
> Fred Isaman wrote:
>> This patch boosts NFS4_PNFS_DEV_MAXSIZE to something slightly more
>> reasonable for the block driver.
>> 
>> FIXME  The code needs to be changed so that, if TOO_SMALL is returned by
>> GETDEVICEINFO, then an appropriately sized GETDEVICEINFO will be sent.
>> 
>> Signed-off-by: Fred Isaman <iisaman at citi.umich.edu>
>> ---
>>  include/linux/nfs4_pnfs.h |    3 ++-
>>  1 files changed, 2 insertions(+), 1 deletions(-)
>> 
>> diff --git a/include/linux/nfs4_pnfs.h b/include/linux/nfs4_pnfs.h
>> index 2aaf526..451afef 100644
>> --- a/include/linux/nfs4_pnfs.h
>> +++ b/include/linux/nfs4_pnfs.h
>> @@ -17,7 +17,8 @@
>>  #include <linux/nfs_page.h>
>>   #define NFS4_PNFS_DEV_MAXNUM 16
>> -#define NFS4_PNFS_DEV_MAXSIZE 128
>> +/* FIXME: This is way too small for block driver */
>> +#define NFS4_PNFS_DEV_MAXSIZE 512
>> 
> 512 is big enough or still too small? Does the comment need to be updated to 
> reflect the FIXME comment at the top? Dean

512 is big enough for testing, and seems adequate as an initial value for 
GETDEVICEINFO4args.gdia_maxcount, but we have to be able to resend with a 
larger value if TOO_SMALL is returned.  (Device encoding can be quite 
large for complex topologies, at minimum containing the disk signature for 
each disk mentioned).

 	Fred


>>   /* Per-layout driver specific registration structure */
>>  struct pnfs_layoutdriver_type {
>>


More information about the pNFS mailing list