Installing Florence rc2 on beta hardware #46, as an FTP install (bootnet.img floppy) from ftp.beta.redhat.com -- text mode portions worked fine and it loaded the ramdisks, but when starting anaconda I immediately got "install exited abnormally -- received signal 11". There was no opportunity presented to capture to floppy disk. 128MB of RAM in this machine. The last lines on VT3: going to insmod ext3.o (path is NULL) looking for USB mouse... The last lines on VT4: <4> raid5: using function: p5_mmx (1050.00 MB/sec) <6> raid5 personality registered as nr 4 VT2 was showing a shell prompt (#) but I couldn't type any commands.
what is beta hardware #46?
oh, sorry - found it.
Did RC1 work with this hw configuration?
Unfortunately I didn't get around to trying RC1 on this particular box. I can certainly try RC1 tonight if it would help; let me know. But what would the answer (RC1 sig11's or doesn't) tell you? It seems to me that (1) nothing should cause anaconda to signal 11, should it? To me that suggests that parameters aren't being checked, null pointers being used, etc. You should at least get an error message. And (2) if it does crash, shouldn't it always give the opportunity for a traceback, so that Red Hat would have something concrete to examine?
This defect considered a show-stopper (MUST-FIX) for Florence GOLD release
OK, got some time to try RC1 on this hardware, and it fails too, in the same way. Furthermore, specifying "nofb", "lowres", and "text" doesn't change it... signal 11 upon starting anaconda, every time. This machine worked fine during RH7 testing, and it's stable with RH7 in daily use.
how much ram?
128MB, as stated above in the initial problem report. :-)
blah. this is all pretty standard hardware - I have no idea why it may be failing.
Agreed. Well, at one point in the past, your QA lab set up a machine similar to mine, to test and reproduce a problem with disk druid partition creation. I'm sure you have all these pieces; as you say, it's all pretty common. About the only thing I haven't provided is my disks: I have three ultra/wide Quantum drives (sda=4.5G, sdb=4.5G, sdc=9G), and then a narrow Plextor CD. (The only IDE is a CD burner and a ZIP drive (ATAPI).) I'm installing 7.1 on parts of sdb and sdc. I'm happy to do whatever I can to help... run a custom kernel and/or initrd with debug code, that sort of thing.
are the the 3 lines above "install exited abnormally"?
There is nothing above the "abnormally" line. The sequence of events: the TUI windows ask for language, keyboard, etc; "loading aic78xx" or something like that appears (loading the SCSI driver); "starting anaconda" or something like that appears very briefly; the screen clears to complete blue, and then "install exited abnormally" appears as the top line of VT1; literally it's in position (1,1) in the upper-left-hand corner. There are a few other messages that appear after that, stair-stepped down the page (like \r without \n is happening). You can see the VT2, VT3, VT4 lines above.
please place the files at http://people.redhat.com/msw/debug on an ext2 formatted floppy disk and boot with "linux updates"
OK, I did that. Unfortunately, here's what I got: [...] Running anaconda - please wait... File "/tmp/updates/anaconda.real", line 515 print "done. ^ SyntaxError: invalid token install exited abnormally sending termination signals... done [...] Looking at it now, clearly I should have just added a double-quote and tried again... but I was very tired and wasn't thinking straight. Sorry about that, I'll try again tonight and let you know; tell me if there are other fixes beyond that one line that I need.
yea - I can't get much further without this run. Sorry about the typo...
OK, well, the output I got: hello, world import installinterface done. make todo install exited abnormally [etc.] I added a few more print statements: Just before method is instantiated, the string value is "ftp://ftp.nordist.net/7.1/rc2/i386" and it gets into the "ftp:// or http://" case to instantiate method. Then, just before the "make todo" line, I think rootPath is "/mnt/sysimage" (I didn't write that one down... sorry, but I do know it's defined, and is "/mnt" something). At that point, I can't continue tracing, because I don't have the source for the ToDo stuff. I'm happy to trace down further in if you'll tell me how to do that. Here is the input I give to the installer between boot and anaconda running, in case the values matter (these are the actual values I'm using), although anaconda crashes with ftp.beta.redhat.com (non-anonymous) too: Choose a Language English Keyboard Type us Installation Method FTP Configure TCP/IP [ ] Use dynamic ... IP address: 10.0.0.2 Netmask: 255.0.0.0 Default gateway (IP): 10.0.0.1 Primary nameserver: 10.0.0.1 FTP Setup FTP site name: ftp.nordist.net Red Hat directory: 7.1/rc2/i386 [ ] Use non-anonymous ftp At that point, ramdisks are loaded, and the AIC 7xxx driver, and then anaconda starts.
I'm going to bet that it is in the kudzu code. I've added a todo.py file to http://people.redhat.com/msw/debug/
This time we didn't even get to hello, world: Running anaconda - please wait... error 2 reading header: cpio: Bad magic cpio failed on (null): (internal) install exited abnormally -- received signal 7 [...] I just dropped todo.py next to anaconda and anaconda.real on the ext2 floppy -- root:root 755 for all three.
it ran out of space expanding stage 2. Try putting only todo.py on the updates disk.
Well, you sure nailed it: instanciate mouse object install exited abnormally FWIW, I have a Microsoft Intellimouse PS/2 (one of the optical ones) in this system.
PS/2 or USB?
This one can do either... I have it connected to the PS/2 port with the USB-to-PS/2 short adapter cable thingie that came with it.
Can you try running the install without the mouse attached or with it attached to usb?
With the mouse completely disconnected (and after a hard reset), the same thing happens (signal 11 after "instanciate mouse"). I don't have any USB on this box, and don't really want to set all that up unless it's the only way for your troubleshooting to continue.
[Testing bugzilla bug 31493 by adding comments to a beta-only bug.]
try booting with "linux nousb" at the boot: prompt. If this doesn't work I'll need to work on getting a boot disk for you with debugging output in the mouse probe routines (written in C, this takes some time.)
Sadly, no, it does the same thing with "linux nousb".
I have a bootnet.img at http://people.redhat.com/msw/debug/bootnet-test.img that will print debugging output as it probes each bus. This should let me narrow it down to a probe function.
Well, we've changed the thing being studied, and now it doesn't crash. (At least, not in the same way.) Here's what I get: Probing OTHER bus Probing PCI bus Probing SBUS bus Probing USB bus Probing SERIAL bus (3-5 second pause at this point) Probing PSAUX bus ... then the "Welcome" dialog pops up. Several more things are behind it on the screen (including PARALLEL, SCSI, IDE, KEYBOARD, ISAPNP) but not all can be read. After the second dialog box, I get: No Drives Found (requester) An error has occurred - no valid devices were found on which to create new filesystems. [...] then it stops and reboots.
# cat /proc/partitions major minor #blocks name [...] # lsmod Module: #pages: Used by: 3c59x 22784 1 keybdev 1968 0 (unused) hid 11072 0 (unused) input 3232 0 [keybdev hid] usb-uhci 21360 0 (unused) usbcore 47408 1 [hid usb-uhci] # Now, if I boot "linux nousb" the only line that shows up under "lsmod" is 3c59x; otherwise, all is the same ("No Drives Found").
Again, no IDE in this machine -- all partitions are SCSI.
oh, sorry - I prepared a wolverine based boot disk. That's why the scsi modules didn't load It looks like the mouse probe crash has been fixed. Do you want a rc2 based boot disk so you can finish a ftp install on this box?
or better yet, try the qa tree posted today.
Unfortunately qa0309 crashes the same way as the original rc2 does (signal 11). So are we back to bus probing (debug) with the Florence RC2 code base?
*grumble grumble*. I'll have a new boot disk soon, with the code from the qa0309 tree.
http://people.redhat.com/msw/debug/bootnet-test2.img
Waiting on feedback based on new bootimage.
(he can only do testing at night... we'll see if we can't get some good amount of testing done tonight)
With bootnet-test2.img (no arguments to LILO) I get: Running anaconda - please wait... error 2 reading header: cpio: Bad magic cpio failed on (null): (internal) install exited abnormally -- recieved signal 7 Now, that's loading the rc2 initrd's over FTP... I'm in the process of pulling the qa0309 initrd's which are probably a better match (and perhaps required with bootnet-test2.img?) but I wanted to get you the early feedback. I'll let you know how the qa0309 initrd's fare.
Same results with the qa0309 initrd's (which I'll begin using from here out unless you ask for something different).
http://people.redhat.com/msw/debug/bootnet-test3.img
See my 2001-03-12 20:08 post; we're past the crash point, and I now have partitions. Do you want the /proc/partitions and lsmod data this time (there's lots of it and I'd have to transcribe it).
that's quite odd. If you could, please continue on with the install and I'd like to try some debugging inside the installed system.
The hitch is, I don't have the full qa0309 here; I brought a ZIP of the tree from disc1 home, leaving out the RPMS and docs directories, just so we could test the early part. I only have modem Internet access from home. :-( It did get all the way through to the point where it would begin loading RPMs. I can do an overnight install of qa0309 from ftp.beta.redhat.com; if I do minimal install it will probably finish by tomorrow morning.
well, the fact that you've gotten past this point with my updated boot disk means that it seems this crash is fixed in the new kudzu.