>>>> Tony.Lill@ajlc.waterloo.on.ca said:
>> IMHO, the massive file corruption that results on a vanilla
>> Linux system is what's not ready for prime time. I've been
>> running the 1.2.2a patch on Linux 2.2.5 for a couple of weeks
>> now, serving 60 HPUX and Solaris boxes without any noticeable
>> problem, whereas before the server was unusable.
Allen> Do you know what was causing the corruption?
Here's a note from USENET, with a little exploit script. I don't know
if the explanation is correct.
I hit this problem trying to compile gcc, the configure script would
fail with bizzarre errors at three or four different points. That is
if I did
rm -rf sparc-solaris2.5.1
mkdir sparc-solaris2.5.1
cd sparc-solaris2.5.1
../configure
several times in a row, it would fail at one of several points.
It would create a file or directory and then not be able to use
it, reporting an error that made me think that it was actually working
on some other filesystem object. Also, files in .. were corrupted, and
I don't think that configure should be doing anything but reading
them.
-- Tony Lill, Tony.Lill@AJLC.Waterloo.ON.CA President, A. J. Lill Consultants fax/data (519) 650 3571 539 Grand Valley Dr., Cambridge, Ont. N3H 2S2 (519) 241 2461 --------------- http://www.ajlc.waterloo.on.ca/ ---------------- "Welcome to All Things UNIX, where if it's not UNIX, it's CRAP!"
>From: mlw <markw@mohawksoft.com> Newsgroups: comp.os.linux.development.system,comp.os.linux.advocacy Subject: nfsd and inode reuse Date: Sat, 10 Apr 1999 12:56:58 +0000 Organization: Posted via RemarQ Communities, Inc. NNTP-Posting-Date: Sat, 10 Apr 1999 12:55:51 GMT X-Trace: 923748951.074.44 IKMOPKJCX4562CD88C usenet1.supernews.com X-Complaints-To: newsabuse@remarQ.com X-Mailer: Mozilla 4.06 [en] (X11; U; Linux 2.2.5-mohawk-1p-nfs-19990403 i586) MIME-Version: 1.0 Content-Transfer-Encoding: 7bit
I have a strange problem with a Solaris machine connected to a Linux server.
It seems that if, in rapid succession, a file is created, deleted, and a new directory is created. In NFS the directory gets the same inode as the previous file. For some reason, the Solaris box wants to think this inode is a file until the nfs client refreshes. As you can imagine this breaks a few scripts. Also, it does not happen 100% of the time (joy ;-( ).
I have a script: >> resuse.sh touch fubar ls -i fubar rm fubar mkdir rabuf ls -id rabuf rmdir rabuf <<
When executed on a local disk, ext2 uses two different inodes. When executed on an NFS volume, it uses the same "inode."
The questions are: Is the reuse of the inode on the NFS mounted volume a problem? or, is it some other level of cache coherency protocol that is breaking down? If it is a Solaris client bug, is there a work around? If it is a Linux server bug, is there a fix?
So, before I start hacking nfsd, maybe it is already fixed ;-)
Also, has anyone else seen this behavior?
If you, can, respond by e-mail. Thanks.
-- Mohawk Software Windows 95, Windows NT, UNIX, Linux. Applications, drivers, support. Take the Mohawk Software Computer Survey at: www.mohawksoft.com- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/