Re: [patch] CONFIG_PREEMPT_REALTIME, 'Fully Preemptible Kernel', VP-2.6.9-rc4-mm1-T4

From: Mark_H_Johnson
Date: Mon Oct 11 2004 - 13:29:55 EST

Next message: Pavel Machek: "Re: suspend-to-RAM [was Re: Totally broken PCI PM calls]"
Previous message: Andre Tomt: "Re: Linux 2.6.9-rc4 - pls test (and no more patches)"
In reply to: Lee Revell: "Re: [patch] CONFIG_PREEMPT_REALTIME, 'Fully Preemptible Kernel',VP-2.6.9-rc4-mm1-T4"
Next in thread: Ingo Molnar: "[patch] VP-2.6.9-rc4-mm1-T5"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

>i've released the -T4 VP patch:
>
>
http://redhat.com/~mingo/voluntary-preempt/voluntary-preempt-2.6.9-rc4-mm1-T4

I would have to say this is "very rough" at this point. I had the following
problems in the build:
[1] kernel/ksyms.c - undefined symbols
[2] kernel/mutex.c - obvious cut / paste problems
[3] XFS has incompatible mutex definition
[4] suspicious warnings
[5] missing symbols for modules
Details at the end.

I booted w/ SMP and the machine threw a lot of error messages about
sleeping in
an invalid context. For example:

include/linux/rwsem.h:43
in_atomic():1 [00010001], irqs_disabled():1
[<c011f0ea>] __might_sleep+0xca/0xe0
[<c01390d4>] rw_mutex_read_lock+0x34/0x50
[<c0122dbd>] profile_hook+0x1d/0x50
[<c0123338>] profile_tick+0x68/0x70
[<c01150ad>] smp_apic_timer_interrupt+0x5d/0xf0
[<c0105820>] default_idle+0x0/0x40
[<c010854a>] apic_timer_interrupt+0x1a/0x20
[<c0105820>] default_idle+0x0/0x40
[<c011007b>] dmi_get_system_info+0xb/0x20
[<c010585a>] default_idle+0x3a/0x40
[<c03b4a4d>] start_kernel+0x19d/0x1e0
[<c03b4440>] unknown_bootoption+0x0/0x180

(somehow managed to stop the scrolling console with the above message
displayed...)

Finally died with a kernel BUG
kernel BUG at kernel/latenc.c:419!
invalid operand: 0000 [#1]
PREEMPT SMP
Modules linked in microcode dm_mod uhci_hcd ext3 jbd
CPU: 1
EIP: 0001:[<00000000>] Not tainted VLI
EFLAGS: c1663f38 (2.6.9-rc4-mm1-VP-T4)
EIP is at 0x0
eax: 00000000 ebx: c1663f54 ecx: c0109a8c edx: c1663f78
esi: 0000000c edi: c1663f28 ebp: c1663f3c esp: c1663f78
ds: 007b es: 07b ss: 4f03 preempt: 00000001
Process swapper (pid: 0, threadinfo=c1662000 task=c165e550)
<0> Kernel panic - not syncing: Attempted to kill the idle task!

Rebooting with num_cpus=1 and appeared to make it farther but then the
console scrolled like crazy and finally said "console shuts up ..."
and the machine appeared to be hung. Could not scroll the window up
or down to see the full message. Had to power off / on to get the
machine back up. Going back to -T3 until I see some fixes.

If the machine managed to record some good data in /var/log/messages
I will send them separately to you for reference.

--Mark H Johnson
<mailto:Mark_H_Johnson@xxxxxxxxxxxx>

Details on build problems / work arounds follow:

[1] ksyms.c - I commented these lines out, to get a complete build, but
there appears to
be code that expects rtc_lock to be defined. See #5.

arch/i386/kernel/i386_ksyms.c:166 error: `rtc_lock' undeclared here (not in
a function)
arch/i386/kernel/i386_ksyms.c:166 error: initializer element is not
constant
arch/i386/kernel/i386_ksyms.c:166 error: (near initialization for
`__ksymtab_rtc_lock.value')
followed by a similar error for atomic_dec_and_lock at line 177.

[2] mutex.c - several symbols were defined twice, fixed by changing the
names to
the functions preceeding them. See lines 108, 201, 213, 297.

[3] XFS compile failed as follows:
CC [M] fs/xfs/quota/xfs_dquot.o
In file included from fs/xfs/linux-2.6/xfs_linux.h:63,
from fs/xfs/xfs.h:35,
from fs/xfs/quota/xfs_dquot.c:33:
fs/xfs/linux-2.6/mutex.h:45: error: conflicting types for `mutex_t'
include/asm/spinlock.h:79: error: previous declaration of `mutex_t'
In file included from fs/xfs/linux-2.6/xfs_linux.h:102,
from fs/xfs/xfs.h:35,
from fs/xfs/quota/xfs_dquot.c:33:
fs/xfs/linux-2.6/xfs_vnode.h:578:30: macro "mutex_lock" requires 2
arguments, but only 1 given
fs/xfs/linux-2.6/xfs_vnode.h:585:30: macro "mutex_lock" requires 2
arguments, but only 1 given
fs/xfs/quota/xfs_dquot.c:1327:23: macro "mutex_lock" requires 2 arguments,
but only 1 given
fs/xfs/quota/xfs_dquot.c:1390:41: macro "mutex_lock" requires 2 arguments,
but only 1 given
fs/xfs/linux-2.6/xfs_vnode.h: In function `vn_flagset':
fs/xfs/linux-2.6/xfs_vnode.h:578: warning: statement with no effect
fs/xfs/linux-2.6/xfs_vnode.h: In function `vn_flagclr':
fs/xfs/linux-2.6/xfs_vnode.h:585: warning: statement with no effect
Turned off XFS in the build.

[4] I considered the following warnings to be "suspicious" but am not sure
if they are really problems or not.

CC security/selinux/ss/policydb.o
fs/dcache.c: In function `prune_dcache':
fs/dcache.c:391: warning: passing arg 1 of `cond_resched_lock' from
incompatible pointer type
CC security/selinux/ss/services.o
CC fs/inode.o
fs/inode.c: In function `invalidate_list':
fs/inode.c:317: warning: passing arg 1 of `cond_resched_lock' from
incompatible pointer type

[5] Several modules had undefined symbols. The messages were...

Kernel: arch/i386/boot/bzImage is ready
*** Warning: "mutex_trylock_bh" [drivers/net/ppp_synctty.ko] undefined!
*** Warning: "del_mtd_partitions" [drivers/mtd/maps/scx200_docflash.ko]
undefined!
*** Warning: "add_mtd_partitions" [drivers/mtd/maps/scx200_docflash.ko]
undefined!
*** Warning: "i2o_msg_in_to_virt" [drivers/message/i2o/i2o_scsi.ko]
undefined!
*** Warning: "i2o_msg_out_to_virt" [drivers/message/i2o/i2o_core.ko]
undefined!
*** Warning: "i2o_msg_in_to_virt" [drivers/message/i2o/i2o_core.ko]
undefined!
*** Warning: "i2o_msg_in_to_virt" [drivers/message/i2o/i2o_block.ko]
undefined!
*** Warning: "rtc_lock" [drivers/char/nvram.ko] undefined!
*** Warning: "rtc_lock" [drivers/char/mwave/mwave.ko] undefined!
*** Warning: "rtc_lock" [drivers/block/floppy.ko] undefined!
...
if [ -r System.map ]; then /sbin/depmod -ae -F System.map -b
/var/tmp/kernel-2.6.9rc4mm1VPT4-root -r 2.6.9-rc4-mm1-VP-T4; fi
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/net/ppp_synctty.ko
needs unknown symbol mutex_trylock_bh
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/mtd/maps/scx200_docflash.ko
needs unknown symbol del_mtd_partitions
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/mtd/maps/scx200_docflash.ko
needs unknown symbol add_mtd_partitions
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/message/i2o/i2o_scsi.ko
needs unknown symbol i2o_msg_in_to_virt
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/message/i2o/i2o_core.ko
needs unknown symbol i2o_msg_in_to_virt
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/message/i2o/i2o_core.ko
needs unknown symbol i2o_msg_out_to_virt
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/message/i2o/i2o_block.ko
needs unknown symbol i2o_msg_in_to_virt
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/char/nvram.ko
needs unknown symbol rtc_lock
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/char/mwave/mwave.ko
needs unknown symbol rtc_lock
WARNING: /var/tmp/kernel-2.6.9
rc4mm1VPT4-root/lib/modules/2.6.9-rc4-mm1-VP-T4/kernel/drivers/block/floppy.ko
needs unknown symbol rtc_lock

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Pavel Machek: "Re: suspend-to-RAM [was Re: Totally broken PCI PM calls]"
Previous message: Andre Tomt: "Re: Linux 2.6.9-rc4 - pls test (and no more patches)"
In reply to: Lee Revell: "Re: [patch] CONFIG_PREEMPT_REALTIME, 'Fully Preemptible Kernel',VP-2.6.9-rc4-mm1-T4"
Next in thread: Ingo Molnar: "[patch] VP-2.6.9-rc4-mm1-T5"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]