I'm running my test program, it uses all available spus to compute via
OpenCL
kernel 3.4-rc1 on a ps3
even on testing spu directly, it crashes
======================================================
[ INFO: possible circular locking dependency detected ]
3.4.0-rc1 #2 Not tainted
-------------------------------------------------------
test/964 is trying to acquire lock:
(&mm->mmap_sem){++++++}, at: [<d0000000005ff430>]
.spufs_ps_fault+0x160/0x210 [spufs]
but task is already holding lock:
(&ctx->state_mutex){+.+...}, at: [<d0000000005ff404>]
.spufs_ps_fault+0x134/0x210 [spufs]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (&ctx->state_mutex){+.+...}:
[<c00000000033974c>] .mutex_lock_interruptible_nested+0x70/0x4a0
[<d000000000600d2c>] .spufs_mem_mmap_fault+0xb0/0x160 [spufs]
[<c0000000000c23f0>] .__do_fault+0x108/0x554
[<c0000000000254d8>] .do_page_fault+0x3c0/0x5b0
[<c0000000000058e8>] handle_page_fault+0x10/0x30
-> #0 (&mm->mmap_sem){++++++}:
[<c000000000084cc8>] .lock_acquire+0x98/0xcc
[<c00000000033ab58>] .down_read+0x34/0x78
[<d0000000005ff430>] .spufs_ps_fault+0x160/0x210 [spufs]
[<c0000000000c23f0>] .__do_fault+0x108/0x554
[<c0000000000254d8>] .do_page_fault+0x3c0/0x5b0
[<c0000000000058e8>] handle_page_fault+0x10/0x30
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&ctx->state_mutex);
lock(&mm->mmap_sem);
lock(&ctx->state_mutex);
lock(&mm->mmap_sem);
*** DEADLOCK ***
1 lock held by test/964:
#0: (&ctx->state_mutex){+.+...}, at: [<d0000000005ff404>]
.spufs_ps_fault+0x134/0x210 [spufs]
stack backtrace:
Call Trace:
[c000000006f6b790] [c000000000010278] .show_stack+0x6c/0x16c (unreliable)
[c000000006f6b840] [c000000000082974] .print_circular_bug+0x2f8/0x330
[c000000006f6b8f0] [c0000000000844f0] .__lock_acquire+0x131c/0x1a5c
[c000000006f6ba40] [c000000000084cc8] .lock_acquire+0x98/0xcc
[c000000006f6bb10] [c00000000033ab58] .down_read+0x34/0x78
[c000000006f6bb90] [d0000000005ff430] .spufs_ps_fault+0x160/0x210 [spufs]
[c000000006f6bc70] [c0000000000c23f0] .__do_fault+0x108/0x554
[c000000006f6bd70] [c0000000000254d8] .do_page_fault+0x3c0/0x5b0
[c000000006f6be30] [c0000000000058e8] handle_page_fault+0x10/0x30
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
