|
From: Ming Z. <mi...@el...> - 2004-06-30 14:27:09
|
i did some tests on .2.2 code and this is my finding.
1) i can create partition(fat32 or ntfs) using ms ini and operate on
them;
2) after i unload the ms ini, i can see and operate on these partition
by loading them from unh ini from another machine.
3) the partition i create and formatted under unh ini can be seen and
operated from ms ini.
4) fdisk /dev/sdb5 -l get this.
Disk /dev/sdb5: 23.9 GB, 23943757824 bytes
255 heads, 63 sectors/track, 2910 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device Boot Start End Blocks Id System
/dev/sdb5p1 1 510 4096543+ b Win95 FAT32
/dev/sdb5p2 511 1147 5116702+ 7 HPFS/NTFS
/dev/sdb5p3 1148 1784 5116702+ 7 HPFS/NTFS
/dev/sdb5p4 1785 2910 9044595 5 Extended
/dev/sdb5p5 1785 2028 1959898+ c Win95 FAT32 (LBA)
/dev/sdb5p6 2029 2910 7084633+ 87 NTFS volume set
so now the problem is how u mount this /dev/sdb5px under linux directly.
since there is no such device node under /dev.
the reason is that once u export a file or partition or a logical volume
from iscsi, it looks like a whole disk drive instead of a partition. so
initiator side will create a new partition table in this partition which
will not be read by normal fdisk utility.
so what is the solution? we need another mechanism to convert a file to
a block device. loop can convert a file to a file system but can not be
a device right?
ming
On Tue, 2004-06-29 at 12:33, FUJITA Tomonoro wrote:
> Hi,
>
> From: "Calvin S. Morrow" <cm...@ve...>
> Subject: [Iscsitarget-devel] Fileio vs Blockio Help
> Date: Mon, 28 Jun 2004 14:39:09 -0600
>
> > I started testing the iSCSI target implementation last week as a
> > possible low-cost alternative in our operations.
>
> Thanks for trying out software.
>
>
> > Fileio mode seems to work great. I have been able to succesfully copy
> > and read files using our Windows initiators without any issues. I can
> > not, however, mount these partitions locally using the linux NTFS tools,
> > as they show bad filesystem layouts, etc.
>
> Do initiators create partitions? Or initiator use disks without
> partitions?
>
>
> > If I format the LVM partitions using the linux NTFS tools, I can not see
> > them with the Windows initiators using fileio mode, however ... if I
> > switch to blockio mode, the target machine can mount the partition and
> > the Windows initiator can use the partition as well.
>
> I don't know why blockio mode works...
>
>
> > blockio mode seems to cause a kernel panic without fail. I can format
> > the partition, attach, etc., but within 1 minute of starting a copy, I
> > get the kernel panic, an assert warning in blockio.c, and cpu register
> > dumps, etc.
> >
> > Is there a way to maintain consistency between being able to do a local
> > mount on the linux target, and still have Windows initiator usage while
> > using fileio mode, or is this impossible? Would I be forced in to using
> > blockio mode to maintain this consistency?
>
> Please don't use blockio mode. It has many problems even if I fix all
> bugs. I implemented it to show how the flexible IO architecture works.
>
> I want to fix blockio mode, but fileio mode is good enough for 99.9
> percent of people. Thus I'm concentrating on other problems.
>
> Anyway, if someone is interested in fixing blockio mode, I'll tell the
> details of the problems.
>
>
> > Just for clarification, I am *NOT* trying to mount the partition while
> > the iSCSI target is running. I just want to have the ability to *DO*
> > both of them, so that I can run various tools, diagnostics, backups,
> > whatever using the linux system without having an iSCSI initiator
> > attached to read the partition.
>
> One possible solution is using 'loopback target'.
>
> First, you install an iSCSI initiator driver (Cisco or UNH) on your
> TARGET host. Next, the target host accesses the disk images that it
> provides to initiators through the initiator driver.
>
> It's not elegant, but it works, I guess.
>
>
> Our target should work with LVM. I don't expect a target to mount the
> volumes. However, it sounds convenient. So I'll dig into this problem
> later (probably this weekend).
>
>
> -------------------------------------------------------
> This SF.Net email sponsored by Black Hat Briefings & Training.
> Attend Black Hat Briefings & Training, Las Vegas July 24-29 -
> digital self defense, top technical experts, no vendor pitches,
> unmatched networking opportunities. Visit www.blackhat.com
> _______________________________________________
> Iscsitarget-devel mailing list
> Isc...@li...
> https://lists.sourceforge.net/lists/listinfo/iscsitarget-devel
--
--------------------------------------------------
| Ming Zhang, PhD. Student
| Dept. of Electrical & Computer Engineering
| College of Engineering
| University of Rhode Island
| Kingston RI. 02881
| e-mail: mingz at ele.uri.edu
| Tel. (401) 874-2293
| Fax. (401) 782-6422
| http://www.ele.uri.edu/~mingz/
--------------------------------------------------
|