New: Possible bug - crash during IDE interface probing.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

           Summary: Possible bug - crash during IDE interface probing.
           Product: IO/Storage
           Version: 2.5
    Kernel Version: 2.6.36.2
          Platform: All
        OS/Version: Linux
              Tree: Mainline
            Status: NEW
          Severity: high
          Priority: P1
         Component: IDE
        AssignedTo: [email protected]
        ReportedBy: [email protected]
        Regression: No

With 2.6.36.1, IDE probing worked. Kernel boots.
With 2.6.36.2, IDE probing crashed kernel.

Hardware: AMD Athlon (x86) 32-bit single-core. SiS5513 IDE interface. PC Chips 810MLR motherboard (from 2002). GCC 4.5.1. [The machine is due for retirement soon. Its replacement may be ordered before 2010 ends.]

When I upgraded (using the same basic configuration file) from .1 to .2, the .2 kernel repeatedly crashed at the point where it displayed on the screen "found SiS5513 IDE interface" and listed it's PCI/IRQ/port numbers, but before it printed any such device (hard drive, cdrom, etc.) found on the interface cable.
Interrupts appear disabled as a keyboard CTRL+ALT+DEL was non-functional, nor
was screen buffer lookback or keyboard LEDs for caps, num, or scroll lock.

Since the kernel initalization never finished, I have no logs. A review of the changelog for .2 did not yield any change to the IDE driver, but some DMA and IRQ stuff was changed. It's quite possible that some kernel configuration setting combinations yield a non-working kernel. However, I saw no differences that explain such.

Therefore, I must ask first: Is there a bug here? If so, with what? If not, then where did my configuration go wrong? Thanks.

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

Borislav Petkov <[email protected]> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |[email protected]

--- Comment #1 from Borislav Petkov <[email protected]> 2010-12-12 12:24:50 --- Hi,

first of all, how did you update from .1 to .2? I'd like to rule out the use of some stale .o objects from the 36.1 build. Can you copy your .config to a safe place, do "make mrproper" and then copy it back (mrproper removes the .config too) and then build 36.2 and try booting.

Then, if it crashes again, can you catch the oops somehow, provided there is one? Either serial console, netconsole or if those are not an option, make a readable photo of the oops and upload it to bugzilla.

Thanks.

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

--- Comment #2 from [email protected] 2010-12-12 21:55:27 --- How do I upgrade? I download the entire source and unpack it into its own directory separate from any prior source. I do NOT patch. Therefore, there are no stale object files. The only component carried from one subversion to another is the configuration file. As for include files installed into /usr/include, I update but I don't permit deletions. Installation script:

----------------------------------------------------
#!/bin/sh
gunzip -c /proc/config.gz >.config
make menuconfig bzImage modules modules_install
make headers_install INSTALL_HDR_PATH=/tmp ; rm find /tmp/include -size 0 (cd /boot ; for x in .config bzImage System.map ; do mv -v $x $x.old ; done) cp -v .config System.map arch/x86/boot/bzImage /boot
cp -R /tmp/include /usr ; rm -r /tmp/include
exec lilo
----------------------------------------------------
As noted, there is no OOPS message. It simply dies:

...
Uniform Multi-Platform E-IDE driver
sis5513 0000:00:00.1: SiS730 ATA 100 (1st gen) controller sis5513 0000:00:00.1: IDE controller (0x1039:0x5513 rev 0xd0) sis5513 0000:00:00.1: IDE port disabled
sis5513 0000:00:00.1: not 100%% native mode: will probe irqs later

    ide0: BM-DMA at 0xff00-0xff07

=> 2.6.36.2 dies here. 2.6.36.1 continues:

Probing IDE interface ide0...
hda: MAXTOR STM3500630A, ATA DISK drive
hdb: TOSHIBA CD-ROM XM-6702B, ATAPI CD/DVD-ROM drive
...
(The above was copied from the 2.6.31.1 syslog.)

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

--- Comment #3 from Borislav Petkov <[email protected]> 2010-12-13 14:13:11 --- Ok,

just wanted to make sure, thanks.

Well, I'm looking through the 290 .36.2 commits and
can't find a strong candidate for this regression. Maybe 3dc40311eaf91cde40f29d799f430dc6a00c2a7c but I can't see how... Can you try reverting it and see if it fixes the issue, per chance?

If that doesn't help, you could still bisect the 290 commits between 36.1 and 36.2.

Feel free to ask in case you need any help.

Thanks.

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

--- Comment #4 from [email protected] 2010-12-17 08:47:49 --- This report is consistent with the problem I initially experienced. I did upgrade my binutils to a 2.21 (from 2.20) version.

http://sourceware.org/bugzilla/show_bug.cgi?id=12327

When I rolled back to a 2.6.36.1 kernel (which I'm running now), I note that I did turn the relocation option OFF. However, it was originally on.

I also note this fix to binutils. I have not installed it to confirm the fix.

http://gcc.gnu.org/ml/gcc/2010-12/msg00397.html

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

--- Comment #5 from Borislav Petkov <[email protected]> 2010-12-17 09:19:10 --- Well,

it's worth trying, if nothing else. There's also a kernel workaround at
http://marc.info/?l=linux-kernel&m=129255612616676 which you could try even with the broken binutils, AFAICT.

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

--- Comment #6 from [email protected] 2010-12-20 21:11:15 --- I installed the new binutils. All the "relocation warnings" have disappeared.

I have not tested the resulting kernel as I have done the upgrades only on my production server and have not rebooted it. I may try this on a spare machine at home, but later (unless someone else can confirm this has disappeared).

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

https://bugzilla.kernel.org/show_bug.cgi?id=24702

[email protected] changed:

What |Removed |Added
----------------------------------------------------------------------------

             Status|NEW                         |RESOLVED
         Resolution|                            |DOCUMENTED

--- Comment #7 from [email protected] 2010-12-23 02:38:59 --- Due to other reasons, I had to reboot the server that had the problem which caused this thread. It is now working.

Resolution: I set it to "Documented" because "Moved" is not an available choice, despite this web site saying so in its help page ("page.cgi?id=fields.html#status"):

"MOVED
"The problem was specific to a related product whose bugs are tracked in another bug database. The bug has been moved to that database."

The related product is binutils. The bug has been fixed over there.

Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.