[FIX] Fix bcm pl011 UART #168
Conversation
FelixSchladt
force-pushed
the
bcm-pl011-uart-fix
branch
from
1e2a8ac
to
407b4d6
Compare
|
Just to clarify, it seem to me that when you enable the receive timeout interrupt and the FIFO, then the serial server should work again. You get an interrupt either when the FIFO has a certain level or - due to the receive timeout - when there is data in the FIFO, but no additional data is received for "a 32-bit period". So we should be able to merge this without breaking thins. I wonder if should still make the interrupt behavior configurable via the |
FelixSchladt
force-pushed
the
bcm-pl011-uart-fix
branch
from
3c0db13
to
0b0c67b
Compare
FelixSchladt
force-pushed
the
bcm-pl011-uart-fix
branch
from
0b0c67b
to
90395bd
Compare
|
The Gitlint and License Checks are stuck, but I can't re-trigger them. I also don't have an option to bring the branch up-to-date, so I guess a manual merge needs to be done. |
Indanz
added
hw-build
|
I will try to rebase it on monday |
FelixSchladt
force-pushed
the
bcm-pl011-uart-fix
branch
from
b177ad8
to
90395bd
Compare
Using the PL011 as uart for data did not work: -Enabling FIFO as no irqs are raised otherwise -Added required interrupts -Fixing incorrect irq disable Signed-off-by: Felix Schladt <[email protected]>
Signed-off-by: Felix Schladt <[email protected]>
FelixSchladt
force-pushed
the
bcm-pl011-uart-fix
branch
from
90395bd
to
e456c24
Compare
|
So I looked at the Broadcom UART documentation, but it doesn't explain what the receive time-out does. Does it trigger all the time or does it only trigger after receiving a character and nothing else arrives? |
As far as I understand, the receive timeout only triggers after a certain period of no incoming data |
Does that mean you get an interrupt all the time? Because that would make it "work", but not in the right way. |
| @@ -227,14 +230,14 @@ static int pl011_uart_configure(ps_chardevice_t *dev) | |||
| * | |||
| */ | |||
| // Enable FIFO | |||
| //pl011_uart_enable_fifo(dev); | |||
| pl011_uart_enable_fifo(dev); | |||
There was a problem hiding this comment.
I'm not convinced that the problem mentioned in the comment is properly addressed. I am afraid that enabling FIFOs is just a work-around for an unfixed bug, probably caused by the original's code assumptions of the current state of the UART or initialisation.
Have you tried fully configuring the UART and setting all registers, and clearing any pending IRQs and errors at init?
There was a problem hiding this comment.
I tested the current state again today and it is working fine.
Sending single bytes is not an issue.
For testing I did setup a simple uart echo server which is interrupt triggered.
There was a problem hiding this comment.
Please print something out at every receive interrupt and check that you don't get any when no input is given.
Or keep a counter and print out the counter value too.
I tried using the existing driver for data transmissions (with a hardware rng) and had the issue that without enabled FIFO, no interrupts are received.
As the note states, enabling the FIFO seems to cause some other trouble with the SerialServer.
I think this should be discussed, what makes the most sense and maybe a solution which fits all can be found.