I do not have a reduced testcase, nor do I know what the real cuplprit is.
kernel compiled with gcc-4_0-branch boot ok, but gcc-mainline fails to boot.
Linux kernel 2.6.14, arch/ppc64/defconfig, do not boot on POWER5.
They hang early in SMP init because a spinlock was unlocked twice.
Enabling CONFIG_DEBUG_SPINLOCK and adding a few extra printk will show this:
....
Linux version 2.6.14 ([EMAIL PROTECTED]) (gcc version 4.1.0 20051102
(experimental)) #61 SMP Wed Nov 2 19:47:05 UTC 2005
....
Memory: 495732k/524288k available (5004k kernel code, 27528k reserved, 1340k
data, 2394k bss, 380k init)
374.78 BogoMIPS (lpj=749568)
Mount-cache hash table entries: 256
softlockup thread 0 started up.
Processor 1 found.
Brought up 2 CPUs
BUG: spinlock already unlocked on CPU#1, migration/1/5
c0 .schedule+0x3fc/0x870c1 .kthread+0xf0/0x170 lock: c0000000142a6100, .magic:
dead4ead, .owner: swapper/1,c0000000004dbc0c,c00000000007b9c0 .owner_cpu: -1
Call Trace:
[c000000000e33c30] [c00000000021d2ec] .spin_bug+0xcc/0x150 (unreliable)
[c000000000e33cc0] [c00000000021d444] ._raw_spin_unlock+0xd4/0xe0
[c000000000e33d50] [c0000000004ddfa4] ._spin_unlock_irq+0xa4/0xe0
[c000000000e33de0] [c0000000004dbc0c] .schedule+0x3fc/0x870
[c000000000e33ee0] [c00000000007b9c0] .kthread+0xf0/0x170
[c000000000e33f90] [c000000000010948] .kernel_thread+0x4c/0x68
kernel BUG in spin_bug at
/home/olaf/kernel/olh/gcc41/linux-2.6.14-olh/lib/spinlock_debug.c:34!
cpu 0x1: Vector: 700 (Program Check) at [c000000000e339b0]
pc: c00000000021d2f0: .spin_bug+0xd0/0x150
lr: c00000000021d2ec: .spin_bug+0xcc/0x150
sp: c000000000e33c30
msr: 8000000000021032
current = 0xc00000000ffd2040
paca = 0xc000000000619000
pid = 5, comm = migration/1
kernel BUG in spin_bug at
/home/olaf/kernel/olh/gcc41/linux-2.6.14-olh/lib/spinlock_debug.c:34!
enter ? for help
1:mon>
In this case, spinlock at c0000000142a6100 is used correctly in a
schedule_tail() and schedule()
call. Later, schedule() starts to use a spinlock at ..142ae100, then suddenly
calls the unlock
with ...142a6100.
I currently do not know if the code in kernel/schedule.c is miscompiled, or if
the breakage is
somewhere else.
If I replace the schedule.o with a gcc40 compiled version, things get further,
but die later
during kernel init due to wrong pointer dereferencing. Perhaps the object files
generated
with gcc-4_0-branch and gcc-mainline are somewhat incompatible.
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 13ff8f28
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 13ff8f28
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 13ff8e90
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 13ff8e90
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 00e17be8
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 00e17be8
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 142a6100
<7>schedule_tail(1552) swapper(6):c0,j4294892318 c0000000142a6100
<7>_spin_unlock_irq(292) swapper(6):c0,j4294892318 l 142a6100
<7>_spin_lock_irq(85) swapper(6):c0,j4294892318 l 00e2c8d0
<7>_spin_unlock_irq(292) swapper(6):c0,j4294892318 l 00e2c8d0
<7>_spin_lock_irq(85) swapper(6):c0,j4294892318 l 142a6100
<7>schedule(3019) swapper(1):c0,j4294892318 c0000000142a6100
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 142a6100
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 00e17be8
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 00e17be8
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 00e17bf8
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 00e17bf8
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 142a6100
<7>schedule_tail(1552) swapper(7):c0,j4294892318 c0000000142a6100
<7>_spin_unlock_irq(292) swapper(7):c0,j4294892318 l 142a6100
<7>_spin_lock_irq(85) swapper(7):c0,j4294892318 l 00e2f9e0
<7>_spin_unlock_irq(292) swapper(7):c0,j4294892318 l 00e2f9e0
<7>_spin_lock_irq(85) swapper(7):c0,j4294892318 l 142a6100
<7>schedule(3019) swapper(1):c0,j4294892318 c0000000142a6100
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 142a6100
<7>_spin_lock_irq(85) swapper(1):c0,j4294892318 l 00e17bf8
<7>_spin_unlock_irq(292) swapper(1):c0,j4294892318 l 00e17bf8
<4>Processor 1 found.
<7>_spin_lock_irq(85) swapper(0):c1,j4294892318 l 142ae100
<6>Brought up 2 CPUs
<7>schedule(3019) swapper(1):c0,j4294892318 c0000000142a6100
<7>_spin_unlock_irq(292) migration/1(5):c1,j4294892318 l 142a6100
<4>BUG: spinlock already unlocked on CPU#1, migration/1/5
<4>c0 .schedule+0x3fc/0x870c1 .kthread+0xf0/0x170 lock: c0000000142a6100,
.magic: dead4ead, .owner: swapper/1,c0000000004dbc0c,c00000000007b9c0
.owner_cpu: -1
<4>Call Trace:
<4>[c000000000e33c30] [c00000000021d2ec] .spin_bug+0xcc/0x150 (unreliable)
<4>[c000000000e33cc0] [c00000000021d444] ._raw_spin_unlock+0xd4/0xe0
<4>[c000000000e33d50] [c0000000004ddfa4] ._spin_unlock_irq+0xa4/0xe0
<4>[c000000000e33de0] [c0000000004dbc0c] .schedule+0x3fc/0x870
<4>[c000000000e33ee0] [c00000000007b9c0] .kthread+0xf0/0x170
<4>[c000000000e33f90] [c000000000010948] .kernel_thread+0x4c/0x68
<2>kernel BUG in spin_bug at
/home/olaf/kernel/olh/gcc41/linux-2.6.14-olh/lib/spinlock_debug.c:34!
1:mon>
--
Summary: gcc-4.1 compiled ppc64 kernels do not boot
Product: gcc
Version: 4.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: olh at suse dot de
GCC build triplet: powerpc64-linux
GCC host triplet: powerpc64-linux
GCC target triplet: powerpc64-linux
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24644
