[gPXE] SANBOOT failed Input/Output Error

[Guo-Fu Tseng]

On Wed, 06 Oct 2010 20:27:22 -0400, Mike Collins wrote
> On Wed, 2010-10-06 at 11:43 +0800, Guo-Fu Tseng wrote:
> > On Mon, 4 Oct 2010 23:46:27 -0400, Michael A. Collins wrote
> > > This is offically driving me mad!  I just setup a iSCSI-SCST box with 
> > > the put_page patches applied and verified that it works by installing 
> > > Fedora 13 from a box to a target.  Now when I go to boot the dang 
> > > thing it doesn't seem to want to.  my iqn is right and whether I type 
> > > it in from the gPXE command-line or set it with option 17, the result 
> > > is the same.  I can see the box connect to my target, but after a few 
> > > back and forths I get the following in my target logs: iscsi-scst: 
> > > __iscsi_state_change:437:***ERROR***: Connection with initiator 
> > > iqn.2000-09.org.etherboot:UNKNOWN unexpectedly closed!
> > > 
> > > This is the only error message that I can make out.  I tried capturing 
> > > the traffic at the target box, but that's even more confusing.  The 
> > > last few packets were the following: target -> client : SCSI: Data In 
> > > LUN: 0x00 (Read Capacity(10) Response Data) client -> target : [PSH, ACK]
> > > client -> target : [FIN, PSH, ACK]
> > > target -> client : [FIN, ACK]
> > > target -> client : [FIN, ACK]
> > > target -> client : [FIN, ACK]
> > > target -> client : [FIN, ACK]
> > > target -> client : [FIN, ACK]
> > > 
> > > So, my question is what the heck is going on here.  Obviously, the 
> > > target must be working or I wouldn't have been able to install the 
> > > client box in the first place.  The partition is bootable and there, 
> > > checked by mounting locally and verifying with fdisk and lvm.  Where 
> > > can I start looking from the gPXE side, is there a way to dump a bunch 
> > > of iscsi logs to the screen during sanboot?  I want to make sure that 
> > > I'm troubleshooting the right piece.  Any help would be great!  I have 
> > > attached a capture file and an exerpt from my target logs. Mike
> > Hi Mike:
> > This issue looks familiar to me.
> > It seems related to the TCP stack fix I worked on before.
> > Could you see if this branch solve the problem:
> > git clone -b tcpfix2 git://git.etherboot.org/scm/people/cooldavid/gpxe.git
> > 
> > Guo-Fu Tseng
> > 
> No change.  Still recieving the same error.
> Mike
Thanks for trying.

So you still got:
> client -> target : [FIN, PSH, ACK]
> target -> client : [FIN, ACK]
> target -> client : [FIN, ACK]
> target -> client : [FIN, ACK]
> target -> client : [FIN, ACK]
> target -> client : [FIN, ACK]
Behavior?

It seem the fix is still missing some check points.
Or something went wrong earlier then gPXE handing
off the control to the OS.

Guo-Fu Tseng