Interrupt scheduling and handling in linux

Question

let's say we are getting 100 interrupts from Net device, 50 interrupts from USB, 25 interrupts from SPI device, 25 interrupts from I2c.

It is coming in sequence as follows 5Net - 4USB - 2SPI -2I2C and the same sequence follows.

The top-level handler can dispatch a device-specific handler to service the interrupt

Now the processor will interrupt the running task as soon as it gets the Net device's interrupt. On completing the Top half of Net device's INterrupt handler, it has to execute the top half of USB and SPI and I2C.

And the same sequence will be followed after completing the 1st set of sequence . When the interrupted task will wake again? Do the interrupted task wait until all the 100 interrupt are serviced by their respective device specific handlers?. How the Interrupts are shared to different cores in case multi-core systems as hundreds of thousands of interrupts will have to be serviced?

As far as I know when executing Interrupt handler the processor will be in interrupt context so that there wont be any context switching. As different ISR will have to service hundreds of thousands of Interrupts, do the processor will always be in interrupt context?

Answer 1

When the interrupted task will wake again?
When interrupts are cleared and the scheduler decides to give this task processor time.

Do the interrupted task wait until all the 100 interrupt are serviced by their respective device specific handlers?
You described only 4 IRQs sources (some net device, usb, spi, i2c). So if all IRQ lines are high and enabled, than the cpus which handle these irqs will switch to specific interrupt handlers. If the interrupt is still triggered after the handler, then the cpu servicing it will branch again and again to the interrupt handler until the interrupt is cleared. On multi-cpu system with 5 cpus, 4 may execute interrupt handlers for your devices simultaneously, while the other one will execute your task. So your task may not be interrupted at all. Or it may wait forever for the cpu, on a single cpu system when the interrupt handler is badly written and never clears the IRQ line.

How the Interrupts are shared to different cores in case multi-core systems as lakhs of interrupts will have to be serviced?
I think it is best explained here: multi-core CPU interrupts .

As different ISR will have to service lakhs of Interrupt, do the processor will always be in interrupt context?
It will stay in interrupt context until the IRQ is enabled and IRQ is triggered. You can just disable the IRQ line and return the cpu to the scheduler, if you need.