mfd: ezx-pcap: Replace mutex_lock with spin_lock
Hello Fuqian Huang,
The patch b65dc4f6b339: "mfd: ezx-pcap: Replace mutex_lock with spin_lock" from Aug 13, 2019, leads to the following Smatch static checker warning:
drivers/spi/spi.c:1666 __spi_pump_transfer_message() warn: sleeping in atomic context
One problematic call tree is:
ezx_pcap_set_bits() <- disables preempt
-> ezx_pcap_putget()
-> spi_sync()
drivers/mfd/ezx-pcap.c
109 int ezx_pcap_set_bits(struct pcap_chip *pcap, u8 reg_num, u32 mask, u32 val) 110 {
111 unsigned long flags; 112 int ret; 113 u32 tmp = PCAP_REGISTER_READ_OP_BIT |
114 (reg_num << PCAP_REGISTER_ADDRESS_SHIFT);
115
116 spin_lock_irqsave(&pcap->io_lock, flags);
This use to be a mutex but it was change to a spinlock to avoid a sleeping in atomic bug.
117 ret = ezx_pcap_putget(pcap, &tmp);
But ezx_pcap_putget() calls spi_sync() which takes a mutex so the fix is not complete.
118 if (ret) 119 goto out_unlock; 120 121 tmp &= (PCAP_REGISTER_VALUE_MASK & ~mask); 122 tmp |= (val & mask) | PCAP_REGISTER_WRITE_OP_BIT | 123 (reg_num << PCAP_REGISTER_ADDRESS_SHIFT); 124 125 ret = ezx_pcap_putget(pcap, &tmp); 126 out_unlock: 127 spin_unlock_irqrestore(&pcap->io_lock, flags); 128 129 return ret; 130 }
regards,
dan carpenter