From eb76e3dc7a93c7cf7d3fe0e7c2852291ec2d1457 Mon Sep 17 00:00:00 2001
From: Colin Xu <colin.xu@intel.com>
Date: Mon, 1 Jun 2020 11:55:56 +0800
Subject: drm/i915/gvt: Use GFP_ATOMIC instead of GFP_KERNEL in atomic context
Git-commit: a291e4fba259a56a6a274c1989997acb6f0bb03a
Patch-mainline: v5.8-rc4
References: jsc#SLE-12680, jsc#SLE-12880, jsc#SLE-12882, jsc#SLE-12883, jsc#SLE-13496, jsc#SLE-15322
GFP_KERNEL flag specifies a normal kernel allocation in which executing
in process context without any locks and can sleep.
mmio_diff takes sometime to finish all the diff compare and it has
locks, continue using GFP_KERNEL will output below trace if LOCKDEP
enabled.
Use GFP_ATOMIC instead.
V2: Rebase.
=====================================================
WARNING: SOFTIRQ-safe -> SOFTIRQ-unsafe lock order detected
5.7.0-rc2 #400 Not tainted
-----------------------------------------------------
is trying to acquire:
ffffffffb47bea20 (fs_reclaim){+.+.}-{0:0}, at: fs_reclaim_acquire.part.0+0x0/0x30
and this task is already holding:
ffff88845b85cc90 (&gvt->scheduler.mmio_context_lock){+.-.}-{2:2}, at: vgpu_mmio_diff_show+0xcf/0x2e0
which would create a new lock dependency:
(&gvt->scheduler.mmio_context_lock){+.-.}-{2:2} -> (fs_reclaim){+.+.}-{0:0}
but this new dependency connects a SOFTIRQ-irq-safe lock:
(&gvt->scheduler.mmio_context_lock){+.-.}-{2:2}
... which became SOFTIRQ-irq-safe at:
lock_acquire+0x175/0x4e0
_raw_spin_lock_irqsave+0x2b/0x40
shadow_context_status_change+0xfe/0x2f0
notifier_call_chain+0x6a/0xa0
__atomic_notifier_call_chain+0x5f/0xf0
execlists_schedule_out+0x42a/0x820
process_csb+0xe7/0x3e0
execlists_submission_tasklet+0x5c/0x1d0
tasklet_action_common.isra.0+0xeb/0x260
__do_softirq+0x11d/0x56f
irq_exit+0xf6/0x100
do_IRQ+0x7f/0x160
ret_from_intr+0x0/0x2a
cpuidle_enter_state+0xcd/0x5b0
cpuidle_enter+0x37/0x60
do_idle+0x337/0x3f0
cpu_startup_entry+0x14/0x20
start_kernel+0x58b/0x5c5
secondary_startup_64+0xa4/0xb0
to a SOFTIRQ-irq-unsafe lock:
(fs_reclaim){+.+.}-{0:0}
... which became SOFTIRQ-irq-unsafe at:
...
lock_acquire+0x175/0x4e0
fs_reclaim_acquire.part.0+0x20/0x30
kmem_cache_alloc_node_trace+0x2e/0x290
alloc_worker+0x2b/0xb0
init_rescuer.part.0+0x17/0xe0
workqueue_init+0x293/0x3bb
kernel_init_freeable+0x149/0x325
kernel_init+0x8/0x116
ret_from_fork+0x3a/0x50
other info that might help us debug this:
Possible interrupt unsafe locking scenario:
CPU0 CPU1
---- ----
lock(fs_reclaim);
local_irq_disable();
lock(&gvt->scheduler.mmio_context_lock);
lock(fs_reclaim);
<Interrupt>
lock(&gvt->scheduler.mmio_context_lock);
*** DEADLOCK ***
3 locks held by cat/1439:
#0: ffff888444a23698 (&p->lock){+.+.}-{3:3}, at: seq_read+0x49/0x680
#1: ffff88845b858068 (&gvt->lock){+.+.}-{3:3}, at: vgpu_mmio_diff_show+0xc7/0x2e0
#2: ffff88845b85cc90 (&gvt->scheduler.mmio_context_lock){+.-.}-{2:2}, at: vgpu_mmio_diff_show+0xcf/0x2e0
the dependencies between SOFTIRQ-irq-safe lock and the holding lock:
-> (&gvt->scheduler.mmio_context_lock){+.-.}-{2:2} ops: 31 {
HARDIRQ-ON-W at:
lock_acquire+0x175/0x4e0
_raw_spin_lock_bh+0x2f/0x40
vgpu_mmio_diff_show+0xcf/0x2e0
seq_read+0x242/0x680
full_proxy_read+0x95/0xc0
vfs_read+0xc2/0x1b0
ksys_read+0xc4/0x160
do_syscall_64+0x63/0x290
entry_SYSCALL_64_after_hwframe+0x49/0xb3
IN-SOFTIRQ-W at:
lock_acquire+0x175/0x4e0
_raw_spin_lock_irqsave+0x2b/0x40
shadow_context_status_change+0xfe/0x2f0
notifier_call_chain+0x6a/0xa0
__atomic_notifier_call_chain+0x5f/0xf0
execlists_schedule_out+0x42a/0x820
process_csb+0xe7/0x3e0
execlists_submission_tasklet+0x5c/0x1d0
tasklet_action_common.isra.0+0xeb/0x260
__do_softirq+0x11d/0x56f
irq_exit+0xf6/0x100
do_IRQ+0x7f/0x160
ret_from_intr+0x0/0x2a
cpuidle_enter_state+0xcd/0x5b0
cpuidle_enter+0x37/0x60
do_idle+0x337/0x3f0
cpu_startup_entry+0x14/0x20
start_kernel+0x58b/0x5c5
secondary_startup_64+0xa4/0xb0
INITIAL USE at:
lock_acquire+0x175/0x4e0
_raw_spin_lock_irqsave+0x2b/0x40
shadow_context_status_change+0xfe/0x2f0
notifier_call_chain+0x6a/0xa0
__atomic_notifier_call_chain+0x5f/0xf0
execlists_schedule_in+0x2c8/0x690
__execlists_submission_tasklet+0x1303/0x1930
execlists_submit_request+0x1e7/0x230
submit_notify+0x105/0x2a4
__i915_sw_fence_complete+0xaa/0x380
__engine_park+0x313/0x5a0
____intel_wakeref_put_last+0x3e/0x90
intel_gt_resume+0x41e/0x440
intel_gt_init+0x283/0xbc0
i915_gem_init+0x197/0x240
i915_driver_probe+0xc2d/0x12e0
i915_pci_probe+0xa2/0x1e0
local_pci_probe+0x6f/0xb0
pci_device_probe+0x171/0x230
really_probe+0x17a/0x380
driver_probe_device+0x70/0xf0
device_driver_attach+0x82/0x90
__driver_attach+0x60/0x100
bus_for_each_dev+0xe4/0x140
bus_add_driver+0x257/0x2a0
driver_register+0xd3/0x150
i915_init+0x6d/0x80
do_one_initcall+0xb8/0x3a0
kernel_init_freeable+0x2b4/0x325
kernel_init+0x8/0x116
ret_from_fork+0x3a/0x50
}
__key.77812+0x0/0x40
... acquired at:
lock_acquire+0x175/0x4e0
fs_reclaim_acquire.part.0+0x20/0x30
kmem_cache_alloc_trace+0x2e/0x260
mmio_diff_handler+0xc0/0x150
intel_gvt_for_each_tracked_mmio+0x7b/0x140
vgpu_mmio_diff_show+0x111/0x2e0
seq_read+0x242/0x680
full_proxy_read+0x95/0xc0
vfs_read+0xc2/0x1b0
ksys_read+0xc4/0x160
do_syscall_64+0x63/0x290
entry_SYSCALL_64_after_hwframe+0x49/0xb3
the dependencies between the lock to be acquired
and SOFTIRQ-irq-unsafe lock:
-> (fs_reclaim){+.+.}-{0:0} ops: 1999031 {
HARDIRQ-ON-W at:
lock_acquire+0x175/0x4e0
fs_reclaim_acquire.part.0+0x20/0x30
kmem_cache_alloc_node_trace+0x2e/0x290
alloc_worker+0x2b/0xb0
init_rescuer.part.0+0x17/0xe0
workqueue_init+0x293/0x3bb
kernel_init_freeable+0x149/0x325
kernel_init+0x8/0x116
ret_from_fork+0x3a/0x50
SOFTIRQ-ON-W at:
lock_acquire+0x175/0x4e0
fs_reclaim_acquire.part.0+0x20/0x30
kmem_cache_alloc_node_trace+0x2e/0x290
alloc_worker+0x2b/0xb0
init_rescuer.part.0+0x17/0xe0
workqueue_init+0x293/0x3bb
kernel_init_freeable+0x149/0x325
kernel_init+0x8/0x116
ret_from_fork+0x3a/0x50
INITIAL USE at:
lock_acquire+0x175/0x4e0
fs_reclaim_acquire.part.0+0x20/0x30
kmem_cache_alloc_node_trace+0x2e/0x290
alloc_worker+0x2b/0xb0
init_rescuer.part.0+0x17/0xe0
workqueue_init+0x293/0x3bb
kernel_init_freeable+0x149/0x325
kernel_init+0x8/0x116
ret_from_fork+0x3a/0x50
}
__fs_reclaim_map+0x0/0x60
... acquired at:
lock_acquire+0x175/0x4e0
fs_reclaim_acquire.part.0+0x20/0x30
kmem_cache_alloc_trace+0x2e/0x260
mmio_diff_handler+0xc0/0x150
intel_gvt_for_each_tracked_mmio+0x7b/0x140
vgpu_mmio_diff_show+0x111/0x2e0
seq_read+0x242/0x680
full_proxy_read+0x95/0xc0
vfs_read+0xc2/0x1b0
ksys_read+0xc4/0x160
do_syscall_64+0x63/0x290
entry_SYSCALL_64_after_hwframe+0x49/0xb3
stack backtrace:
CPU: 5 PID: 1439 Comm: cat Not tainted 5.7.0-rc2 #400
Hardware name: Intel(R) Client Systems NUC8i7BEH/NUC8BEB, BIOS BECFL357.86A.0056.2018.1128.1717 11/28/2018
Call Trace:
dump_stack+0x97/0xe0
check_irq_usage.cold+0x428/0x434
? check_usage_forwards+0x2c0/0x2c0
? class_equal+0x11/0x20
? __bfs+0xd2/0x2d0
? in_any_class_list+0xa0/0xa0
? check_path+0x22/0x40
? check_noncircular+0x150/0x2b0
? print_circular_bug.isra.0+0x1b0/0x1b0
? mark_lock+0x13d/0xc50
? __lock_acquire+0x1e32/0x39b0
__lock_acquire+0x1e32/0x39b0
? timerqueue_add+0xc1/0x130
? register_lock_class+0xa60/0xa60
? mark_lock+0x13d/0xc50
lock_acquire+0x175/0x4e0
? __zone_pcp_update+0x80/0x80
? check_flags.part.0+0x210/0x210
? mark_held_locks+0x65/0x90
? _raw_spin_unlock_irqrestore+0x32/0x40
? lockdep_hardirqs_on+0x190/0x290
? fwtable_read32+0x163/0x480
? mmio_diff_handler+0xc0/0x150
fs_reclaim_acquire.part.0+0x20/0x30
? __zone_pcp_update+0x80/0x80
kmem_cache_alloc_trace+0x2e/0x260
mmio_diff_handler+0xc0/0x150
? vgpu_mmio_diff_open+0x30/0x30
intel_gvt_for_each_tracked_mmio+0x7b/0x140
vgpu_mmio_diff_show+0x111/0x2e0
? mmio_diff_handler+0x150/0x150
? rcu_read_lock_sched_held+0xa0/0xb0
? rcu_read_lock_bh_held+0xc0/0xc0
? kasan_unpoison_shadow+0x33/0x40
? __kasan_kmalloc.constprop.0+0xc2/0xd0
seq_read+0x242/0x680
? debugfs_locked_down.isra.0+0x70/0x70
full_proxy_read+0x95/0xc0
vfs_read+0xc2/0x1b0
ksys_read+0xc4/0x160
? kernel_write+0xb0/0xb0
? mark_held_locks+0x24/0x90
do_syscall_64+0x63/0x290
entry_SYSCALL_64_after_hwframe+0x49/0xb3
RIP: 0033:0x7ffbe3e6efb2
Code: c0 e9 c2 fe ff ff 50 48 8d 3d ca cb 0a 00 e8 f5 19 02 00 0f 1f 44 00 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 0f 05 <48> 3d 00 f0 ff ff 77 56 c3 0f 1f 44 00 00 48 83 ec 28 48 89 54 24
RSP: 002b:00007ffd021c08a8 EFLAGS: 00000246 ORIG_RAX: 0000000000000000
RAX: ffffffffffffffda RBX: 0000000000020000 RCX: 00007ffbe3e6efb2
RDX: 0000000000020000 RSI: 00007ffbe34cd000 RDI: 0000000000000003
RBP: 00007ffbe34cd000 R08: 00007ffbe34cc010 R09: 0000000000000000
R10: 0000000000000022 R11: 0000000000000246 R12: 0000562b6f0a11f0
R13: 0000000000000003 R14: 0000000000020000 R15: 0000000000020000
------------[ cut here ]------------
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Colin Xu <colin.xu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200601035556.19999-1-colin.xu@intel.com
Signed-off-by: Patrik Jakobsson <pjakobsson@suse.de>
---
drivers/gpu/drm/i915/gvt/debugfs.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/gvt/debugfs.c b/drivers/gpu/drm/i915/gvt/debugfs.c
index ec47d4114554..62e6a14ad58e 100644
--- a/drivers/gpu/drm/i915/gvt/debugfs.c
+++ b/drivers/gpu/drm/i915/gvt/debugfs.c
@@ -66,7 +66,7 @@ static inline int mmio_diff_handler(struct intel_gvt *gvt,
vreg = vgpu_vreg(param->vgpu, offset);
if (preg != vreg) {
- node = kmalloc(sizeof(*node), GFP_KERNEL);
+ node = kmalloc(sizeof(*node), GFP_ATOMIC);
if (!node)
return -ENOMEM;
--
2.28.0