ppc32 lockups with 2.6

ppc32 lockups with 2.6

[Christian Kujau]

moin,

i have problems with my ppc32 (PReP) machine and 2.6.0-test* kernels.
since this is some kind of home "server", i did not test any ealier
versions, but 2.6.0-test{7|8|9}. all show the same behaviour, somewhere
during the bootup, the machine just freezes. i found out, that upon
configuring my thernet devices the lockups occur.

so, i have 2 network cards, both are working under 2.4 with the 3c59x
(eth0) and tulip (eth1, it's an on-board 21140).

loading the drivers by using insmod / modprobe succeeds, the driver
registers, some messages are printed.

e.g.

Linux Tulip driver version 1.1.13 (May 11, 2002)
PCI: Enabling device 0000:00:04.0 (0000 -> 0003)
tulip0:  EEPROM default media type Autosense.
tulip0:  Index #0 - Media AUI (#2) described by a 21140 non-MII (0) block.
tulip0:  Index #1 - Media MII (#11) described by a 21140 MII PHY (1) block.
tulip0:  MII transceiver #8 config 3100 status 786b advertising 01e1.
eth0: Digital DS21140 Tulip rev 34 at 0x1800, 08:00:3E:29:0B:B6, IRQ 9.

or

3c59x: Donald Becker and others. www.scyld.com/network/vortex.html
See Documentation/networking/vortex.txt
0000:00:06.0: 3Com PCI 3c905C Tornado at 0x1000. Vers LK1.1.19
  00:01:02:f1:83:37, IRQ 11
   product code 464a rev 00.13 date 11-12-00
   Internal config register is 1800000, transceivers 0xa.
   8K byte-wide RAM 5:3 Rx:Tx split, autoselect/Autonegotiate interface.
   MII transceiver found at address 24, status 782d.
   Enabling bus-master transmits and whole-frame receives.
0000:00:06.0: scatter/gather enabled. h/w checksums enabled

"ifconfig eth0" gives then an unconfigured eth0, ok. but using "ifconfig
eth0 192.168.1.1" results in a lockup. no oops printed, no SysReq anymore.

i tried with the "tulip" modul first, and could "strace" the ifconfig,
until it freezes.

    http://nerdbynature.de/bits/sheep/network/ifc-tulip.log

the 3c59x module however acted somehow differently: upon using "ifconfig
eth0 192.168.1.1" the screen filled with strange chars in some pattern,
i've never seen before on a linux box (no, i was clean :-)) (ok, the
pattern is probably not important, but it's on [1] anyway) then, the
system locked up too.

as you can see (http://nerdbynature.de/bits/sheep/network/config) i have
then compiled some debug options into the kernel, i loaded the 3c59x
module with "debug=6". also, i want to say that this is not my first
build of a 2.6 kernel, i have a i386 running pretty stable with 2.6 for
weeks now; some time ago i used to have an Alpha (AXP) with 2.5.x
running 24x7.

please let me know, i have forgotten something or where to look further.
i will be glad to help with running tests and the like.

i don't suspect hw errors here, because both cards are happy working
under 2.4.

some details about the PrEP machine:
[1] http://nerdbynature.de/bits/

Thank you,
Christian.

--
BOFH excuse #396:

Mail server hit by UniSpammer.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian]

as nobody seems to answer,

yes, i know, if the problem does not occur to other people, it's not
worth fixing it. and PReP machines seem to be some kind of "exotic"
(read: outdated) anyways. but i don't have the skills fixing it :-(

so, if at least someone could point me to something i might have
forgotten to look at, or how to get an oops message out of this here.
would be nice if it turns out to be a user-error only.

i also filed a bug in bugzilla:
http://bugzilla.kernel.org/show_bug.cgi?id=1494

Thanks for reading,
Christian.

--
BOFH excuse #128:

Power Company having EMP problems with their reactor


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[linas]

On Thu, Nov 06, 2003 at 10:34:26PM +0100, Christian wrote:
>
> as nobody seems to answer,
>
> yes, i know, if the problem does not occur to other people, it's not
> worth fixing it. and PReP machines seem to be some kind of "exotic"
> (read: outdated) anyways. but i don't have the skills fixing it :-(
>
> so, if at least someone could point me to something i might have
> forgotten to look at, or how to get an oops message out of this here.
> would be nice if it turns out to be a user-error only.

I'd suggest getting the KDB kernel patches, and then poking around to
see what the system was doing when it locked up.  What were the stack traces?
does it always lock up in the same routine?  What line of code was it
executiing when it did that?

--linas


>
> i also filed a bug in bugzilla:
> http://bugzilla.kernel.org/show_bug.cgi?id=1494
>
> Thanks for reading,
> Christian.
>
> --
> BOFH excuse #128:
>
> Power Company having EMP problems with their reactor
>
>
>

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Tom Rini]

On Mon, Nov 10, 2003 at 12:50:44PM -0600, linas@austin.ibm.com wrote:
>
> On Thu, Nov 06, 2003 at 10:34:26PM +0100, Christian wrote:
> >
> > as nobody seems to answer,
> >
> > yes, i know, if the problem does not occur to other people, it's not
> > worth fixing it. and PReP machines seem to be some kind of "exotic"
> > (read: outdated) anyways. but i don't have the skills fixing it :-(
> >
> > so, if at least someone could point me to something i might have
> > forgotten to look at, or how to get an oops message out of this here.
> > would be nice if it turns out to be a user-error only.
>
> I'd suggest getting the KDB kernel patches, and then poking around to
> see what the system was doing when it locked up.  What were the stack traces?
> does it always lock up in the same routine?  What line of code was it
> executiing when it did that?

It should be easier to get KGDB working here actually.  I know in the
not too distant past I've hooked KGDB up to my Powerstack for some
quick testing.

--
Tom Rini
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian]

linas@austin.ibm.com wrote:
> I'd suggest getting the KDB kernel patches, and then poking around to
> see what the system was doing when it locked up.

kdb, hm, i'll see if i can handle this. would some output from gdb help too?


> What were the stack traces? does it always lock up in the same routine?

the debugger will show, as nothing else showed up and even SysReq was
not working anymore.

> What line of code was it
> executiing when it did that?

last snippet of the "strace ifconfig eth1 192.168.1.1":

ioctl(4, 0x8916, 0x7ffffd78)            = -1 ENODEV (No such device)
dup(2)                                  = 5
fcntl64(5, F_GETFL)                     = 0x2 (flags O_RDWR)
fstat64(5, {st_mode=S_IFCHR|0600, st_rdev=makedev(5, 1), ...}) = 0
ioctl(5, TCGETS, {B38400 opost isig icanon echo ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0)
= 0x30016000
_llseek(5, 0, 0x7ffffb68, SEEK_CUR)     = -1 ESPIPE (Illegal seek)
write(5, "SIOCSIFADDR: No such device\n", 28) = 28
close(5)                                = 0
munmap(0x30016000, 4096)                = 0
ioctl(4, 0x8913, 0x7ffffc98)            = -1 ENODEV (No such device)
write(2, "eth1: ERROR while getting interf"..., 58) = 58
exit(-1)                                = ?

Thanks,
Christian.

--
BOFH excuse #231:

We had to turn off that service to comply with the CDA Bill.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[linas]

On Mon, Nov 10, 2003 at 11:39:14PM +0100, Christian wrote:
> linas@austin.ibm.com wrote:
> > I'd suggest getting the KDB kernel patches, and then poking around to
> > see what the system was doing when it locked up.
>
> kdb, hm, i'll see if i can handle this. would some output from gdb help too?

kdb is a built-in kernel debugger. gdb is useless for this case,
although kgdb could be another alternative.

> > What were the stack traces? does it always lock up in the same routine?
>
> the debugger will show, as nothing else showed up and even SysReq was
> not working anymore.

Hmm,  that's bad ... does sysreq work before the lockup?

> > What line of code was it
> > executiing when it did that?

No, I meant in the kernel, the output of the kdb 'bt' command, which
shows what the kernel was doing when kdb was entereed.  But if sysreq
really doesn't work, then theres a chance that kdb won't either, and
that means ... hmm...

--linas

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian]

linas@austin.ibm.com wrote:
>
> kdb is a built-in kernel debugger. gdb is useless for this case,
> although kgdb could be another alternative.

yes, i thought so. i'm still a bit confused if "kdb" and "kgdb".

as http://kgdb.sourceforge.net/whichdebugger.html points out, kdb would
be the better choice here, but i don't see any "kdb" within the kernel
config nor does it seem to be a user space-tool. anyway, i have

CONFIG_DEBUG_KERNEL=y
CONFIG_DEBUG_SLAB=y
CONFIG_MAGIC_SYSRQ=y
CONFIG_KGDB=y
CONFIG_KGDB_TTYS1=y
CONFIG_XMON=y
CONFIG_DEBUG_INFO=y
CONFIG_BOOTX_TEXT=y

in my kernel now (still compiling), and will try to setup kgdb tomorrow
(read: today :-))

Thank you for your time,
Christian.
--
BOFH excuse #410:

Electrical conduits in machine room are melting.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Tom Rini]

On Tue, Nov 11, 2003 at 01:25:41AM +0100, Christian wrote:
>
> linas@austin.ibm.com wrote:
> >
> >kdb is a built-in kernel debugger. gdb is useless for this case,
> >although kgdb could be another alternative.
>
> yes, i thought so. i'm still a bit confused if "kdb" and "kgdb".
>
> as http://kgdb.sourceforge.net/whichdebugger.html points out, kdb would
> be the better choice here, but i don't see any "kdb" within the kernel
> config nor does it seem to be a user space-tool. anyway, i have
>
> CONFIG_DEBUG_KERNEL=y
> CONFIG_DEBUG_SLAB=y
> CONFIG_MAGIC_SYSRQ=y
> CONFIG_KGDB=y
> CONFIG_KGDB_TTYS1=y
> CONFIG_XMON=y
> CONFIG_DEBUG_INFO=y
> CONFIG_BOOTX_TEXT=y

Note that with XMON and KGDB both enabled, you'll get xmon instead,
which again I _think_ works on PReP (but it was long ago when I tested)
and here I think you'll be better off with KGDB.

--
Tom Rini
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian Kujau]

Tom Rini wrote:
>>
>>CONFIG_DEBUG_KERNEL=y
>>CONFIG_DEBUG_SLAB=y
>>CONFIG_MAGIC_SYSRQ=y
>>CONFIG_KGDB=y
>>CONFIG_KGDB_TTYS1=y
>>CONFIG_XMON=y
>>CONFIG_DEBUG_INFO=y
>>CONFIG_BOOTX_TEXT=y
>
> Note that with XMON and KGDB both enabled, you'll get xmon instead,
> which again I _think_ works on PReP (but it was long ago when I tested)
> and here I think you'll be better off with KGDB.

so little time, so much to compile, so slow processors :-)
and my compile fails with CONFIG_KGDB=y. i have unpacked a fresh
2.6.0-test9, did "make oldconfig", then only enabled KGDB, but compile
bails out with:

evil@sheep:/usr/src/linux-2.6$ make bzImage
make[1]: `arch/ppc/kernel/asm-offsets.s' is up to date.
   CHK     include/asm-ppc/offsets.h
   CHK     include/linux/compile.h
   UPD     include/linux/compile.h
   CC      init/version.o
   LD      init/built-in.o
   GEN     .version
   CHK     include/linux/compile.h
   UPD     include/linux/compile.h
   CC      init/version.o
   LD      init/built-in.o
   LD      .tmp_vmlinux1
arch/ppc/kernel/built-in.o(.init.text+0x81c): In function `setup_arch':
arch/ppc/kernel/setup.c:635: undefined reference to `kgdb_map_scc'
make: *** [.tmp_vmlinux1] Error 1
evil@sheep:/usr/src/linux-2.6$


so, it seems to depend on "SCC_CONSOLE", but this is only available when
  "MPC8260 CPM Options" is set. but this "MPC8260" is a different
platform. ok, when i choose a different platform, say PRPMC800, i still
can select 6xx/7xx/74xx/8260 as cpu-type (i have 604). no, i don't seem
to understand this, and in-kernel documentation of the different
platforms is not so clear :-(

the .config is on http://www.nerdbynature.de/bits/sheep/network/config_new

anyway, i'm about to setup a cross-compile environment on my AMD-K7
(i386), will see what it gives. oh, i noticed, that on i386 we don't
have CONFIG_KGDB at all, right?

thanks for reading,
Christian.

PS: oh, played around with .config (yes, the above config-new is still
correct) i got:

evil@sheep:/usr/src/linux-2.6$ make bzImage
make[1]: `arch/ppc/kernel/asm-offsets.s' is up to date.
   CHK     include/asm-ppc/offsets.h
   CHK     include/linux/compile.h
   GEN     .version
   CHK     include/linux/compile.h
   UPD     include/linux/compile.h
   CC      init/version.o
   LD      init/built-in.o
   LD      .tmp_vmlinux1
arch/ppc/kernel/built-in.o(.text+0xa628): In function `getpacket':
arch/ppc/kernel/ppc-stub.c:350: undefined reference to `getDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa640):arch/ppc/kernel/ppc-stub.c:359:
undefined reference to `getDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa6b0):arch/ppc/kernel/ppc-stub.c:373:
undefined reference to `getDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa6c0):arch/ppc/kernel/ppc-stub.c:374:
undefined reference to `getDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa6e4):arch/ppc/kernel/ppc-stub.c:376:
undefined reference to `putDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa6f0):arch/ppc/kernel/ppc-stub.c:378:
undefined reference to `putDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa704):arch/ppc/kernel/ppc-stub.c:381:
undefined reference to `putDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa70c):arch/ppc/kernel/ppc-stub.c:382:
undefined reference to `putDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa768): In function `putpacket':
arch/ppc/kernel/ppc-stub.c:402: undefined reference to `putDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa780):arch/ppc/kernel/ppc-stub.c:412:
more undefined references to `putDebugChar' follow
arch/ppc/kernel/built-in.o(.text+0xa79c): In function `putpacket':
arch/ppc/kernel/ppc-stub.c:415: undefined reference to `getDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa7c8):arch/ppc/kernel/ppc-stub.c:407:
undefined reference to `putDebugChar'
arch/ppc/kernel/built-in.o(.text+0xa9bc): In function `handle_exception':
arch/ppc/kernel/ppc-stub.c:589: undefined reference to `kgdb_interruptible'
arch/ppc/kernel/built-in.o(.text+0xae18):arch/ppc/kernel/ppc-stub.c:781:
undefined reference to `kgdb_interruptible'
arch/ppc/kernel/built-in.o(.init.text+0x81c): In function `setup_arch':
arch/ppc/kernel/setup.c:635: undefined reference to `kgdb_map_scc'
arch/ppc/platforms/built-in.o(.init.text+0x930): In function
`pmac_setup_arch':
arch/ppc/platforms/pmac_setup.c:301: undefined reference to `zs_kgdb_hook'
make: *** [.tmp_vmlinux1] Error 1
--
BOFH excuse #380:

Operators killed when huge stack of backup tapes fell over.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Tom Rini]

On Thu, Nov 13, 2003 at 01:59:03AM +0100, Christian Kujau wrote:
> Tom Rini wrote:
> >>
> >>CONFIG_DEBUG_KERNEL=y
> >>CONFIG_DEBUG_SLAB=y
> >>CONFIG_MAGIC_SYSRQ=y
> >>CONFIG_KGDB=y
> >>CONFIG_KGDB_TTYS1=y
> >>CONFIG_XMON=y
> >>CONFIG_DEBUG_INFO=y
> >>CONFIG_BOOTX_TEXT=y
> >
> >Note that with XMON and KGDB both enabled, you'll get xmon instead,
> >which again I _think_ works on PReP (but it was long ago when I tested)
> >and here I think you'll be better off with KGDB.
>
> so little time, so much to compile, so slow processors :-)
> and my compile fails with CONFIG_KGDB=y. i have unpacked a fresh
> 2.6.0-test9, did "make oldconfig", then only enabled KGDB, but compile
> bails out with:
>
> evil@sheep:/usr/src/linux-2.6$ make bzImage
> make[1]: `arch/ppc/kernel/asm-offsets.s' is up to date.
>   CHK     include/asm-ppc/offsets.h
>   CHK     include/linux/compile.h
>   UPD     include/linux/compile.h
>   CC      init/version.o
>   LD      init/built-in.o
>   GEN     .version
>   CHK     include/linux/compile.h
>   UPD     include/linux/compile.h
>   CC      init/version.o
>   LD      init/built-in.o
>   LD      .tmp_vmlinux1
> arch/ppc/kernel/built-in.o(.init.text+0x81c): In function `setup_arch':
> arch/ppc/kernel/setup.c:635: undefined reference to `kgdb_map_scc'
> make: *** [.tmp_vmlinux1] Error 1
> evil@sheep:/usr/src/linux-2.6$
>
>
> so, it seems to depend on "SCC_CONSOLE", but this is only available when
>  "MPC8260 CPM Options" is set. but this "MPC8260" is a different
> platform. ok, when i choose a different platform, say PRPMC800, i still
> can select 6xx/7xx/74xx/8260 as cpu-type (i have 604). no, i don't seem
> to understand this, and in-kernel documentation of the different
> platforms is not so clear :-(
>
> the .config is on http://www.nerdbynature.de/bits/sheep/network/config_new

Ok, it's been a while since anyone tried KGDB it seems.  The following
is an untested (yet) patch that makes it compile and maybe work:

===== arch/ppc/Kconfig 1.46 vs edited =====
--- 1.46/arch/ppc/Kconfig	Mon Oct  6 03:07:31 2003
+++ edited/arch/ppc/Kconfig	Thu Nov 13 10:23:41 2003
@@ -585,7 +585,7 @@

 config PPC_GEN550
 	bool
-	depends on SANDPOINT
+	depends on SANDPOINT || PPC_PREP
 	default y

 config PPC_PMAC
===== arch/ppc/platforms/pmac_setup.c 1.34 vs edited =====
--- 1.34/arch/ppc/platforms/pmac_setup.c	Fri Sep 26 16:31:59 2003
+++ edited/arch/ppc/platforms/pmac_setup.c	Thu Nov 13 10:41:44 2003
@@ -297,8 +297,10 @@
 			ppc_override_l2cr_value, (ppc_override_l2cr_value & 0x80000000)
 				? "enabled" : "disabled");

+#if 0
 #ifdef CONFIG_KGDB
 	zs_kgdb_hook(0);
+#endif
 #endif

 #ifdef CONFIG_ADB_CUDA
===== arch/ppc/platforms/prep_setup.c 1.41 vs edited =====
--- 1.41/arch/ppc/platforms/prep_setup.c	Fri Sep 12 09:26:54 2003
+++ edited/arch/ppc/platforms/prep_setup.c	Thu Nov 13 10:36:38 2003
@@ -39,6 +39,8 @@
 #include <linux/ide.h>
 #include <linux/seq_file.h>
 #include <linux/root_dev.h>
+#include <linux/serial.h>
+#include <linux/serial_core.h>

 #include <asm/sections.h>
 #include <asm/mmu.h>
@@ -58,6 +60,7 @@
 #include <asm/i8259.h>
 #include <asm/open_pic.h>
 #include <asm/pci-bridge.h>
+#include <asm/serial.h>

 unsigned char ucSystemType;
 unsigned char ucBoardRev;
@@ -78,6 +81,7 @@

 extern void prep_find_bridges(void);
 extern char saved_command_line[];
+extern void gen550_init(int i, struct uart_port *serial_req);

 int _prep_type;

@@ -684,6 +688,40 @@
 #endif /* CONFIG_PREP_RESIDUAL */
 }

+#if defined(CONFIG_SERIAL_8250) && \
+	(defined(CONFIG_KGDB) || defined(CONFIG_SERIAL_TEXT_DEBUG))
+static void __init
+prep_early_serial_map(void)
+{
+	struct uart_port serial_req;
+
+	/* Setup serial port access */
+	memset(&serial_req, 0, sizeof(serial_req));
+	serial_req.uartclk = UART_CLK;
+	serial_req.irq = 4;
+	serial_req.flags = STD_COM_FLAGS;
+	serial_req.iotype = SERIAL_IO_PORT;
+	serial_req.iobase = 0x3F8;
+
+	gen550_init(0, &serial_req);
+
+	if (early_serial_setup(&serial_req) != 0)
+		printk(KERN_ERR "Early serial init of port 0 failed\n");
+
+	/* Assume early_serial_setup() doesn't modify serial_req */
+	serial_req.line = 1;
+	serial_req.irq = 3; /* XXXX */
+	serial_req.iobase = 0x2F8;
+
+	gen550_init(1, &serial_req);
+
+	if (early_serial_setup(&serial_req) != 0)
+		printk(KERN_ERR "Early serial init of port 1 failed\n");
+}
+#endif
+
+
+
 static void __init
 prep_setup_arch(void)
 {
@@ -1176,6 +1214,14 @@
 	}

 	ppc_md.setup_io_mappings = prep_map_io;
+
+#if defined(CONFIG_SERIAL_8250) && \
+	(defined(CONFIG_KGDB) || defined(CONFIG_SERIAL_TEXT_DEBUG))
+	prep_early_serial_map();
+#ifdef CONFIG_SERIAL_TEXT_DEBUG
+	ppc_md.progress = gen550_progress;
+#endif
+#endif

 #if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_IDE_MODULE)
 	ppc_ide_md.default_irq = prep_ide_default_irq;
===== arch/ppc/syslib/gen550_dbg.c 1.1 vs edited =====
--- 1.1/arch/ppc/syslib/gen550_dbg.c	Tue Jul  1 08:34:08 2003
+++ edited/arch/ppc/syslib/gen550_dbg.c	Thu Nov 13 10:38:48 2003
@@ -17,6 +17,8 @@
  */

 #include <linux/config.h>
+#include <linux/types.h>
+#include <linux/serial.h>
 #include <linux/tty.h>		/* For linux/serial_core.h */
 #include <linux/serial_core.h>
 #include <linux/serialP.h>
===== include/asm-ppc/pc_serial.h 1.2 vs edited =====
--- 1.2/include/asm-ppc/pc_serial.h	Sun Sep 15 21:52:05 2002
+++ edited/include/asm-ppc/pc_serial.h	Thu Nov 13 10:32:27 2003
@@ -19,6 +19,7 @@
  * megabits/second; but this requires the faster clock.
  */
 #define BASE_BAUD ( 1843200 / 16 )
+#define UART_CLK	1843200

 #ifdef CONFIG_SERIAL_MANY_PORTS
 #define RS_TABLE_SIZE  64


> anyway, i'm about to setup a cross-compile environment on my AMD-K7
> (i386), will see what it gives. oh, i noticed, that on i386 we don't
> have CONFIG_KGDB at all, right?

Correct.

--
Tom Rini
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian Kujau]

Tom Rini wrote:
>
> Ok, it's been a while since anyone tried KGDB it seems.  The following
> is an untested (yet) patch that makes it compile and maybe work:
>
> ===== arch/ppc/Kconfig 1.46 vs edited =====
[...]


ok, i could have done the stuff in Kconfig though, next time i will do
so at once. well, compiling wnt fine on the ppc now (my i386 cross
compile is still pending), but the things with the serial console will
have to wait now. hm, i'm not online at the weekend, so don't expect
much results these days. thank you very much for checking back to this
issue, i (and hopefully some other PreP people out there) really
appreciate this.

Christian.

PS: oh, will this patch reach mainline anytime soon?
--
BOFH excuse #433:

error: one bad user found in front of screen


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Tom Rini]

On Fri, Nov 14, 2003 at 12:56:04AM +0100, Christian Kujau wrote:

> Tom Rini wrote:
> >
> >Ok, it's been a while since anyone tried KGDB it seems.  The following
> >is an untested (yet) patch that makes it compile and maybe work:
> >
> >===== arch/ppc/Kconfig 1.46 vs edited =====
> [...]
>
>
> ok, i could have done the stuff in Kconfig though, next time i will do
> so at once. well, compiling wnt fine on the ppc now (my i386 cross
> compile is still pending), but the things with the serial console will
> have to wait now. hm, i'm not online at the weekend, so don't expect
> much results these days. thank you very much for checking back to this
> issue, i (and hopefully some other PreP people out there) really
> appreciate this.
>
> Christian.
>
> PS: oh, will this patch reach mainline anytime soon?

Probably once Andrew Morton takes over again, at least some parts of the
patch will go in, an the rest possibly in 2.6.1, depending on how much
he wants to let arches sync up.

--
Tom Rini
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian Kujau]

hi,

2.6.0-test9 with kgdb compiled fine then, but i did not get a serial
cable yet.

i took another approach, i wanted to see since when the lockups occur. i
was testing 2.5 kernels on i386 and alpha, but i've started to test 2.6
on ppc pretty late. so, what i see is, that i can boot to 2.6.0-test4
(former kernel version down to 2.5.72 did not even compile cleanly).
i can set up eth0 and i can watch "ifconfig eth0" then. "ping
192.168.1.10" (eth0) is working, but routing seems broken. (i have to
show the details later on).

"ping 192.168.1.11" fails and print a lot to the syslog as you can see
in line 581 in:

http://nerdbynature.de/bits/sheep/network/2.6.0-test4/messages.gz

during bootup i had to use SYSREQ-E to quit a process, that wanted to
use the ethernet-device (snort?)
(SYSREQ-K should have been enough perhaps).

here's the config for this 2.6.0-test4 kernel
http://nerdbynature.de/bits/sheep/network/2.6.0-test4/config

i will go on with 2.6.0-tes5 now....

Thank you,
Christian.
--
BOFH excuse #229:

wrong polarity of neutron flow


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian Kujau]

Christian Kujau wrote:
> i will go on with 2.6.0-tes5 now....

no strange error messages anymore with -test5, but still no routing.
i someone wants to have a look:

http://nerdbynature.de/bits/

(i've re-orderd the links, sorry. the "my PReP machine (ppc32) doesn't
like the new kernel" are the right ones.)

Thanks,
Christian.

--
BOFH excuse #179:

multicasts on broken packets


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Christian Kujau]

hi again (and sorry to contact you directly, Tom),

i did not make it with kgdb, but rather compiled kernels over and over
again this week. my gcc-3.3.2 won't compile on a faster i386, so
crosscompiling was not an option :-(

i went down to 2.5.56 and found that even this kernel has the same
symptoms. in short: upon using "tulip", the machine freezes, "3c59x" is
loaded, but not working. (no routing).

i went even down to 2.5.30 to find out, that this kernel is *working*!
both tulip and 3c59x get loaded and are working. i did not test long,
don't know if it will freeze after a day or so, but i think not, as the
described freezes with newer kernel occur instantly.

the thing is, 2.5.31 ist not compiling, nor is 2.5.3x, except this very
2.5.30. same config, same compiler. 2.5.30 compile over and over again,
newer kernels (2.5.31...up to 2.5.46) *do not compile*. so it's pretty
hard to find out *when* the symptoms occur, to dig into this very patch
and finally find out *why*.

i *hope* 2.5.31 or at least 2.5.32 is the bad one. i am curently looking
through the 2.5.31 patch and sorted out ppc-related stuff, to make it
compile, then shuffle a few bits back in so, that it will compile and i
hopefully get my freeze. yes, you can see that this kind of working is
rather...er..crap. but i can't do better at the moment :-(

thanks for reading,
Christian.
--
BOFH excuse #27:

radiosity depletion


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ppc32 lockups with 2.6

[Tom Rini]

On Fri, Nov 21, 2003 at 04:25:52PM +0100, Christian Kujau wrote:

> hi again (and sorry to contact you directly, Tom),
>
> i did not make it with kgdb, but rather compiled kernels over and over
> again this week. my gcc-3.3.2 won't compile on a faster i386, so
> crosscompiling was not an option :-(
>
> i went down to 2.5.56 and found that even this kernel has the same
> symptoms. in short: upon using "tulip", the machine freezes, "3c59x" is
> loaded, but not working. (no routing).
>
> i went even down to 2.5.30 to find out, that this kernel is *working*!
> both tulip and 3c59x get loaded and are working. i did not test long,
> don't know if it will freeze after a day or so, but i think not, as the
> described freezes with newer kernel occur instantly.
>
> the thing is, 2.5.31 ist not compiling, nor is 2.5.3x, except this very
> 2.5.30. same config, same compiler. 2.5.30 compile over and over again,
> newer kernels (2.5.31...up to 2.5.46) *do not compile*. so it's pretty
> hard to find out *when* the symptoms occur, to dig into this very patch
> and finally find out *why*.
>
> i *hope* 2.5.31 or at least 2.5.32 is the bad one. i am curently looking
> through the 2.5.31 patch and sorted out ppc-related stuff, to make it
> compile, then shuffle a few bits back in so, that it will compile and i
> hopefully get my freeze. yes, you can see that this kind of working is
> rather...er..crap. but i can't do better at the moment :-(

I'd suggest talking to:
NETWORK DEVICE DRIVERS
P:      Andrew Morton
M:      akpm@digeo.com
P:      Jeff Garzik
M:      jgarzik@pobox.com
L:      linux-net@vger.kernel.org

specifically, as they probably know what changed in terms of network
code in between 2.5.30 and 2.5.46 and might even be able to give you
diffs to apply to 2.6.0-testCurrent.

--
Tom Rini
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/