IRQ questions
I am trying to hunt down what I believe is an IRQ problem associated with my serial driver.
I guess I am using old style interrupts.
In 2.4.0-test 5 the file arch/mips/kernel/irq.c used to be built. Under 2.4.3 it no longer is. However, it appears it was renamed to old_irg.c and a new irq.c was created. I also noticed that an i8259.c files was added at some point. However, it does not seem to match the code that was in irq.c.
Are there any notes available that explain how to convert from old style IRQ's to new?
What are we suppose to do with the new irq.c which is not being used?
I have a 2.4.3 kernel booting. I copied the old arch/mips/kernel/irq.c to my target directory and changed
a few other things to get everything to compile. As long as I do not try to use a serial port everything seems to be working. I booted with a frame buffer and started X. The mouse and keyboard worked so some of my IRQ's appear to be working.
When I start the serial port things seem to go fine untile it trys to write the port at which point it starts running very slowly. Which is what make me think the kernel is being overcome with interrupts.
Thanks for any thoughts or advice.
Scott
Quoted 5 lines · Scott A McConnell
In 2.4.0-test 5 the file
arch/mips/kernel/irq.cused to be built. Under 2.4.3 it no longer is. However, it appears it was renamed toold_irg.cand a newirq.cwas created. I also noticed that ani8259.cfiles was added at some point. However, it does not seem to match the code that was inirq.c.
Not entirely, yet. The idea is that a board can register a structure containing information on how to handle an interrupt with the new irq.c So a board which has PC-style i8259 PICs could call init_i8259_irqs which in turn would register handlers for the interrupts 0 - 15 which would be handle in i8259; handler for other interrupts controllers could be registered also.
Quoted 3 lines
Are there any notes available that explain how to convert from old style IRQ's to new?
What are we suppose to do with the newirq.cwhich is not being used?
The IRQ changes should be considered work in progress; until the new interrupt code you should just follow the example given by the many existing implementations.
Quoted 2 lines
I have a 2.4.3 kernel booting. I copied the old
arch/mips/kernel/irq.cto my target directory and changed
One valid solution ... Still. We want to eleminate all this code duplication for no good reason.
Quoted 8 lines
a few other things to get everything to compile. As long as I do not try to use a serial port everything seems to be working. I booted with a frame buffer and started X. The mouse and keyboard worked so some of my IRQ's appear to be working.
When I start the serial port things seem to go fine untile it trys to write the port at which point it starts running very slowly. Which is what make me think the kernel is being overcome with interrupts.
Or you don't receive interrupts at all.
Ralf
Quoted 5 lines · Ralf Baechle
Quoted 2 lines
I have a 2.4.3 kernel booting. I copied the old
arch/mips/kernel/irq.cto my target directory and changedOne valid solution ... Still. We want to eleminate all this code duplication for no good reason.
Would Rotten_IRQ have done the same thing?
Could you name an arch in the cvs distribution that uses the new style IRQ's
Thanks,
Scott
Quoted 7 lines · Scott A McConnell
Quoted 5 lines
Quoted 2 lines
I have a 2.4.3 kernel booting. I copied the old
arch/mips/kernel/irq.cto my target directory and changedOne valid solution ... Still. We want to eleminate all this code duplication for no good reason.
Would
Rotten_IRQhave done the same thing?
Absolutely not. The rotten IRQ thing gives you the old irq.c which assumes more or less x86 centric interrupt handling, that is a 2 PICs and doesn't get SMP right and leaves us a ton of structural problems. Don't use, it's destilled evil ;-)
Quoted 1 line
Could you name an arch in the cvs distribution that uses the new style IRQ's
Interrupt handling is an per architecture thing; currently i386 is coming closest to the new design of our future interrupts.
Ralf
A message that is not in the archive (sent off-list or not yet received): [email protected]
Jun & Ralf,
Thanks for the info.
I will dig into the code that Jun provided.
Jun, I noticed the new/old time code with the last update. It is on my TODO list. Thanks for taking the time to point it out and writing the README.
Scott