A good and a bad news

[Vincent ROQUETA]

Le mercredi 21 mars 2007 17:19, Talpey, Thomas a écrit :
> Have you tried replacing the sync with a sleep? Your code isn't
> checking the return code from read(), and it seems that it will
> continue to attempt read()s for some time after reaching eof,
> depending on whether this /mnt/nfs/ro file is smaller than 20MB.
This not the case : the test file is really larger than 20MB (64GB )

> If the file is larger, then perhaps readahead is happening.
That is a better explanation, but I wan to be sure there will not be any 
corruption. 

> BTW, what exactly is this test trying to do? It prints the time
> but it also seems to be checksumming the data it reads, and it
> is reading one block more than the size. I think!
Well, this program is a part of a larger test set to compare access to very 
large data (+16GB) upon different pattern access. This is one of the most 
basic access : sequential read.

The checksum is done to make all tests for all patterns in the same 
conditions, and to be sure to *really* access all data in the wanted field : 
for example, mmap will not really load data if there is not explicitly 
acceded.

Vincent




> Tom.
>
> At 09:46 AM 3/21/2007, Vincent ROQUETA wrote:
> >Hello all,
> >
> >Two news.
> >First the bad :
> >The umount problem I have seen on NFSv4 exists also on v3
> >
> >The good one :
> >If I do a sync on v3, the problem disappear - Well, consider that is a
> > read only test, closing the file it use.
> >
> >
> >mount -t nfs 129.183.140.144:/Backup/ /mnt/nfs/ ; read_s /mnt/nfs/ro
> >20000000 ; umount /mnt/nfs
> >-> Return : /mnt/nfs: device is busy
> >
> >mount -t nfs 129.183.140.144:/Backup/ /mnt/nfs/ ; read_s /mnt/nfs/ro
> >20000000 ; sync ; umount /mnt/nfs
> >-> correct umount
> >
> >I join the read_s test program. (Read with an initial seek )
> >
> >Have a nice day ...
> >Vincent
> >
> >
> >_______________________________________________
> >NFSv4 mailing list
> >NFSv4 at linux-nfs.org
> >http://linux-nfs.org/cgi-bin/mailman/listinfo/nfsv4