[PATCH 4.14 07/85] btrfs: reschedule when cloning lots of extents

From: Greg Kroah-Hartman
Date: Tue Nov 17 2020 - 08:15:16 EST


From: Johannes Thumshirn <johannes.thumshirn@xxxxxxx>

[ Upstream commit 6b613cc97f0ace77f92f7bc112b8f6ad3f52baf8 ]

We have several occurrences of a soft lockup from fstest's generic/175
testcase, which look more or less like this one:

watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [xfs_io:10030]
Kernel panic - not syncing: softlockup: hung tasks
CPU: 0 PID: 10030 Comm: xfs_io Tainted: G L 5.9.0-rc5+ #768
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.13.0-0-gf21b5a4-rebuilt.opensuse.org 04/01/2014
Call Trace:
<IRQ>
dump_stack+0x77/0xa0
panic+0xfa/0x2cb
watchdog_timer_fn.cold+0x85/0xa5
? lockup_detector_update_enable+0x50/0x50
__hrtimer_run_queues+0x99/0x4c0
? recalibrate_cpu_khz+0x10/0x10
hrtimer_run_queues+0x9f/0xb0
update_process_times+0x28/0x80
tick_handle_periodic+0x1b/0x60
__sysvec_apic_timer_interrupt+0x76/0x210
asm_call_on_stack+0x12/0x20
</IRQ>
sysvec_apic_timer_interrupt+0x7f/0x90
asm_sysvec_apic_timer_interrupt+0x12/0x20
RIP: 0010:btrfs_tree_unlock+0x91/0x1a0 [btrfs]
RSP: 0018:ffffc90007123a58 EFLAGS: 00000282
RAX: ffff8881cea2fbe0 RBX: ffff8881cea2fbe0 RCX: 0000000000000000
RDX: ffff8881d23fd200 RSI: ffffffff82045220 RDI: ffff8881cea2fba0
RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000032
R10: 0000160000000000 R11: 0000000000001000 R12: 0000000000001000
R13: ffff8882357fd5b0 R14: ffff88816fa76e70 R15: ffff8881cea2fad0
? btrfs_tree_unlock+0x15b/0x1a0 [btrfs]
btrfs_release_path+0x67/0x80 [btrfs]
btrfs_insert_replace_extent+0x177/0x2c0 [btrfs]
btrfs_replace_file_extents+0x472/0x7c0 [btrfs]
btrfs_clone+0x9ba/0xbd0 [btrfs]
btrfs_clone_files.isra.0+0xeb/0x140 [btrfs]
? file_update_time+0xcd/0x120
btrfs_remap_file_range+0x322/0x3b0 [btrfs]
do_clone_file_range+0xb7/0x1e0
vfs_clone_file_range+0x30/0xa0
ioctl_file_clone+0x8a/0xc0
do_vfs_ioctl+0x5b2/0x6f0
__x64_sys_ioctl+0x37/0xa0
do_syscall_64+0x33/0x40
entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7f87977fc247
RSP: 002b:00007ffd51a2f6d8 EFLAGS: 00000206 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f87977fc247
RDX: 00007ffd51a2f710 RSI: 000000004020940d RDI: 0000000000000003
RBP: 0000000000000004 R08: 00007ffd51a79080 R09: 0000000000000000
R10: 00005621f11352f2 R11: 0000000000000206 R12: 0000000000000000
R13: 0000000000000000 R14: 00005621f128b958 R15: 0000000080000000
Kernel Offset: disabled
---[ end Kernel panic - not syncing: softlockup: hung tasks ]---

All of these lockup reports have the call chain btrfs_clone_files() ->
btrfs_clone() in common. btrfs_clone_files() calls btrfs_clone() with
both source and destination extents locked and loops over the source
extent to create the clones.

Conditionally reschedule in the btrfs_clone() loop, to give some time back
to other processes.

CC: stable@xxxxxxxxxxxxxxx # 4.4+
Reviewed-by: Josef Bacik <josef@xxxxxxxxxxxxxx>
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@xxxxxxx>
Reviewed-by: David Sterba <dsterba@xxxxxxxx>
Signed-off-by: David Sterba <dsterba@xxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
fs/btrfs/ioctl.c | 2 ++
1 file changed, 2 insertions(+)

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 56123ce3b9f0e..d3f76e3efd44c 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -3863,6 +3863,8 @@ process_slot:
ret = -EINTR;
goto out;
}
+
+ cond_resched();
}
ret = 0;

--
2.27.0