Bug 217563
Summary: | Can't install using VNC on zSeries | ||||||
---|---|---|---|---|---|---|---|
Product: | Red Hat Enterprise Linux 5 | Reporter: | Bastien Nocera <bnocera> | ||||
Component: | anaconda | Assignee: | David Cantrell <dcantrell> | ||||
Status: | CLOSED ERRATA | QA Contact: | |||||
Severity: | medium | Docs Contact: | |||||
Priority: | medium | ||||||
Version: | 5.0 | CC: | bhinson, jlaska, jmbezekiel, tao | ||||
Target Milestone: | --- | ||||||
Target Release: | --- | ||||||
Hardware: | s390 | ||||||
OS: | Linux | ||||||
Whiteboard: | |||||||
Fixed In Version: | RHBA-2007-0644 | Doc Type: | Bug Fix | ||||
Doc Text: | Story Points: | --- | |||||
Clone Of: | Environment: | ||||||
Last Closed: | 2007-11-07 17:18:50 UTC | Type: | --- | ||||
Regression: | --- | Mount Type: | --- | ||||
Documentation: | --- | CRM: | |||||
Verified Versions: | Category: | --- | |||||
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||
Cloudforms Team: | --- | Target Upstream Version: | |||||
Embargoed: | |||||||
Bug Depends On: | 278261 | ||||||
Bug Blocks: | 227613, 234654, 303961 | ||||||
Attachments: |
|
Description
Bastien Nocera
2006-11-28 17:05:26 UTC
This request was evaluated by Red Hat Product Management for inclusion in a Red Hat Enterprise Linux major release. Product Management has requested further review of this request by Red Hat Engineering, for potential inclusion in a Red Hat Enterprise Linux Major release. This request is not yet committed for inclusion. Does this happen with the final beta2? Oh, wait -- this is an ftp install with ~ 256 megs of RAM. So you're getting bumped to minstg2.img, which doesn't have vnc. We shouldn't be trying to do vnc if only text mode is available (which is the case here) This is still present in RHEL-5 nightly trees. If we don't have X stuff available on the stage2 image, we shouldn't give the user the option of running VNC. Setting the RHEL-5.1 flag to ? Tested HTTP VNC installs on a s390x guest with 768Mb ... VNC started without error. Testing HTTP VNC installs on a s390x guest with mem=256m yields: The VNC server is now running. Please connect to lister.z900.redhat.com:1 to begin the install... Press <enter> for a shell Starting graphical installation... Exception starting GUI installer: No module named gtk Traceback (most recent call last): File "/usr/bin/anaconda", line 875, in ? anaconda.setInstallInterface(opts.display_mode) File "/usr/bin/anaconda", line 514, in setInstallInterface isys.vtActivate (1) File "/usr/lib/anaconda/isys.py", line 831, in vtActivate _isys.vtActivate (num) SystemError: (22, 'Invalid argument') Do we need to adjust our documentation for the mem limits? Yeah, this is a documentation issue or release notes issue. Will 512MB work as the lower bound? I think it should ... but let me pull in jjarvis for IBM's take on this. Perhaps there is a bound and it's just not clear. Looking at comment 4 and comment 5, this doesn't seem to be a documentation issue. How complex would it be to add a memory size check before starting the graphical install? If we're using minstg2, we really shouldn't attempt to start VNC. If my understanding of comments 4 and 5 are incorrect please let me know. I added the check for an exectuable Xvnc back in April, which is what the fix for this bug was originally. The traceback now shows: Exception starting GUI installer: No module named gtk So we're missing a lot of libraries and/or python modules, so we're probably still in minstg2 but somehow have Xvnc. I'm not sure. OK, looking in to it more, what's happening is that the memory check _is_ working and the graphical install fails. We catch the failure in anaconda and move you to text install. But on s390x, our vtActivate() function fails, most likely because we can't perform that ioctl() on s390x (chalk up another one for the mainframe being unique). Here's the code: try: from gui import InstallInterface except Exception, e: stdoutLog.error("Exception starting GUI installer: %s" %(e,)) if flags.test: sys.exit(1) # if we're not going to really go into GUI mode, we need to get # back to vc1 where the text install is going to pop up. if not x_already_set: isys.vtActivate (1) stdoutLog.warning("GUI installer startup failed, falling back to text mode.") display_mode = 't' if 'DISPLAY' in os.environ.keys(): del os.environ['DISPLAY'] time.sleep(2) I think we need to just not run vtActivate() for s390x. There are no VTs on that platform at that point and the device isn't really present, so that makes sense that it would fail. Making a patch. OK, here's my proposed patch: Index: isys/isys.py =================================================================== RCS file: /usr/local/CVS/anaconda/isys/isys.py,v retrieving revision 1.193.4.16 diff -u -p -r1.193.4.16 isys.py --- isys/isys.py 10 Aug 2007 17:40:34 -0000 1.193.4.16 +++ isys/isys.py 29 Aug 2007 20:00:26 -0000 @@ -828,6 +828,8 @@ def driveIsIscsi(device): return False def vtActivate (num): + if rhpl.getArch() == "s390": + return _isys.vtActivate (num) def isPsudoTTY (fd): Simple fix to not run vtActivate() on s390 since we don't have VTs. Patched committed. Will be in anaconda-11.1.2.69-1. Testing against RHEL5.1-Server-20070904.nightly (anaconda-11.1.2.71-1) 19:57:07 WARNING : URLINSTALL falling back to non-GUI stage2 due to insufficient RAM However the main attached console shows ... 20:01:58 INFO : Running anaconda script /usr/bin/anaconda^M 20:01:58 DEBUG : adding extraArg --vnc^M 20:01:59 INFO : using only installclass Red Hat Enterprise Linux^M 20:01:59 INFO : Red Hat Enterprise Linux is only installclass, using it^M 20:01:59 INFO : Display mode = g^M 20:01:59 INFO : Method = http://porkchop.redhat.com/nightly/RHEL5.1-Server-20070904.nightly/tree-s390x^M 20:01:59 WARNING : no floppy devices found but we'll try fd0 anyway^M 20:02:02 INFO : Starting VNC...^M 20:02:02 WARNING : ^M ^M WARNING!!! VNC server running with NO PASSWORD!^M You can use the vncpassword=<password> boot option^M if you would like to secure the server.^M ^M ^M 20:02:02 INFO : The VNC server is now running.^M 20:02:02 INFO : Please connect to lister.z900.redhat.com:1 to begin the install...^M 20:02:02 INFO : Started mini-wm^M 20:02:02 INFO : Starting graphical installation...^M 20:02:02 ERROR : Exception starting GUI installer: No module named gtk^M 20:02:02 WARNING : GUI installer startup failed, falling back to text mode.^M 20:02:05 WARNING : Unable to find temp path, going to use ramfs path^M 20:02:05 INFO : Detected 256M of memory^M 20:02:05 INFO : Swap attempt of 256M to 512M^M 20:02:05 WARNING : step installtype does not exist^M 20:02:05 WARNING : step complete does not exist^M 20:02:05 INFO : moving (1) to step welcome^M I then am presented with stage#2 over ssh ... but keyboard inputs do not respond ┌──────┤ Red Hat Enterprise Linux Server ├──────┠│ │ │ Welcome to Red Hat Enterprise Linux Server! │ │ │ │ │ │ ┌────┠│ │ │ OK │ │ │ └────┘ │ │ │ │ │ └───────────────────────────────────────────────┘ Created attachment 186581 [details]
anaconda.log (RHEL5.1-Server-20070904.nightly)
With the patch from bug #278261, the problem described here goes away. VNC installs work fine. Possible related bug -- attempted a VNC install of Fedora 7.90 (F8T1) tonight on an i386 platform, similar symptoms. Is this already fixed in the 7.91 install media? VERIFIED original problem on RHEL5.1-Server-20070920.1 19:10:37 WARNING : URLINSTALL falling back to non-GUI stage2 due to insufficient RAM 19:10:37 INFO : transferring http://porkchop.redhat.com//rel-eng/RHEL5.1-Server-20070920.1/tree-s390x/images/min 19:10:41 DEBUG : copied 44269568 bytes to /tmp/ramfs/minstg2.img (complete) 19:10:41 INFO : mntloop loop0 on /mnt/runtime as /tmp/ramfs/minstg2.img fd is 9 <snip> 19:10:43 INFO : Running anaconda script /usr/bin/anaconda 19:10:43 DEBUG : adding extraArg --vnc 19:10:46 INFO : using only installclass Red Hat Enterprise Linux 19:10:46 INFO : Red Hat Enterprise Linux is only installclass, using it 19:10:46 INFO : Display mode = g 19:10:46 INFO : Method = http://porkchop.redhat.com/rel-eng/RHEL5.1-Server-20070920.1/tree-s390x 19:10:46 WARNING : no floppy devices found but we'll try fd0 anyway 19:10:50 INFO : Starting VNC... 19:10:50 INFO : ip of eth0 is None 19:10:50 WARNING : WARNING!!! VNC server running with NO PASSWORD! You can use the vncpassword=<password> boot option if you would like to secure the server. 19:10:50 INFO : The VNC server is now running. 19:10:50 INFO : Please connect to localhost.localdomain:1 to begin the install... 19:10:50 INFO : Started mini-wm 19:10:50 INFO : Starting graphical installation... 19:10:51 ERROR : Exception starting GUI installer: No module named gtk 19:10:51 WARNING : GUI installer startup failed, falling back to text mode. 19:10:54 WARNING : Unable to find temp path, going to use ramfs path 19:10:54 INFO : Detected 256M of memory 19:10:54 INFO : Swap attempt of 256M to 512M 19:10:54 WARNING : step installtype does not exist 19:10:54 WARNING : step complete does not exist 19:10:54 INFO : moving (1) to step welcome 19:11:16 INFO : moving (1) to step regkey The keyboard/input problem remains ... I'm able to F12 to the regkey step, but then stuck there. I don't seem to see a text-mode cursor. So I don't get the traceback anymore, and I imagine one could do a kickstart install in this environment now. I'm unable to perform an attended installation in a mem=256m s390x zVM guest. dcantrell: Should we close out this issue since the original traceback isn't occuring, and open a new issue to track the keyboard focus issue? Moving to VERIFIED ... will track the keyboard input issue as a separate request (see bug#303961) An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2007-0644.html |