When booting up preempt rt kernel on stratix10 platform, there is
below calltrace:
BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:969
in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 1, name: swapper/0
Preemption disabled at:
[<ffff8000100b25b0>] __setup_irq+0xe0/0x730
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.10.78-rt54-yocto-preempt-rt #1
Hardware name: SoCFPGA Stratix 10 SoCDK (DT)
Call trace:
......
___might_sleep+0x144/0x180
rt_spin_lock+0x34/0x9c
regmap_lock_spinlock+0x24/0x40
regmap_write+0x48/0x84
a10_eccmgr_irq_unmask+0x34/0x40
......
altr_edac_a10_probe+0x16c/0x2b0
platform_drv_probe+0x60/0xb4
......
ret_from_fork+0x10/0x38
Because regmap_write is invoked under preemption disabled status, and
might trigger sleep.
Recently, the commit 67021f25d952("regmap: teach regmap to use raw
spinlocks if requested in the config ") add an option for regmap to use
raw spinlock. So, enable raw spinlock in preempt-rt kernel to avoid the
might sleep issue.
Note: The commit 67021f25d952 is only in kernel v5.15, not included in
earlier versions. So, if intend to fix this issue in earlier preempt-rt
kernel, it is need to backport the 2 patches together, otherwise there
will be building issue.
Signed-off-by: Meng Li <[email protected]>
---
v3:
- remove #ifdef CONFIG_PREEMPT_RT
v2:
- improve the patch title.
- improve the commit log. Clear which commit is depended by this patch.
---
drivers/mfd/altera-sysmgr.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/mfd/altera-sysmgr.c b/drivers/mfd/altera-sysmgr.c
index 5d3715a28b28..ac350b4ca486 100644
--- a/drivers/mfd/altera-sysmgr.c
+++ b/drivers/mfd/altera-sysmgr.c
@@ -83,6 +83,7 @@ static struct regmap_config altr_sysmgr_regmap_cfg = {
.fast_io = true,
.use_single_read = true,
.use_single_write = true,
+ .use_raw_spinlock = true,
};
/**
--
2.17.1