On 8/7/20 10:01 AM, Jiaxun Yang wrote:
>> I'm testing a PCIe card to two uarts (WCH CH382L compatible,
>>           https://item.jd.com/48547850173.html) on a loongson ls2k1000 board.
>>
>> With the card inserted the kernel detects the two new uarts:
>>
>> [    2.060884] pci 0000:00:04.1: EHCI: unrecognized capability ff
>> [    2.066721] pci 0000:00:04.1: EHCI: capability loop?
>> [    2.066745] pci 0000:00:04.1: quirk_usb_early_handoff+0x0/0xa60 took 358764 usecs
>> [    2.074542] pcieport 0000:00:09.0: PME: Signaling with IRQ 2
>> [    2.080651] pcieport 0000:00:0a.0: PME: Signaling with IRQ 3
>> [    2.086648] pcieport 0000:00:0b.0: PME: Signaling with IRQ 4
>> [    2.092616] pcieport 0000:00:0c.0: PME: Signaling with IRQ 5
>> [    2.098579] pcieport 0000:00:0d.0: PME: Signaling with IRQ 6
>> [    2.104557] pcieport 0000:00:0e.0: PME: Signaling with IRQ 7
>> [    2.113728] Serial: 8250/16550 driver, 4 ports, IRQ sharing enabled
>> [    2.120891] serial 0000:06:00.0: limiting MRRS to 256
>> [    2.126314] 0000:06:00.0: ttyS0 at I/O 0xc0 (irq = 7, base_baud = 115200) is a XR16850
>> [    2.134618] 0000:06:00.0: ttyS1 at I/O 0xc8 (irq = 7, base_baud = 115200) is a XR16850
>> [    2.143130] 1fe00000.serial: ttyS2 at MMIO 0x1fe00000 (irq = 1, base_baud = 7812500) is a 16550A
>> [    2.151946] printk: console [ttyS2] enabled
>> [    2.160255] printk: bootconsole [early0] disabled
>> [    2.170112] [drm] radeon kernel modesetting enabled.
>>
>> However a write to the serial fails:
>> / # echo "hello" >/dev/ttyS0
>> [ 9369.631915] serial 0000:06:00.0: LSR safety check engaged!
>> sh: write error: Input/output error
> 
> I assume you're using my out-of-tree fork to support LS2K.
> I messed with Logic PIO in that fork.
> I'll fix it in mainline, sorry for the issue...

Do you have a tree that we can test?

Marc