On Mon, Feb 26, 2018 at 11:05:29PM +0200, Nikolay Nikolov wrote:
On Sun, 2018-02-25 at 18:33 -0500, Kevin O'Connor wrote:
> On Sun, Feb 25, 2018 at 10:38:09PM +0200, Nikolay Nikolov wrote:
> > This ensures that we safely receive the IRQ even if the floppy
> > controller
> > manages to be faster than our code and sends the IRQ before our
> > code enters
> > floppy_wait_irq(). Although unlikely, this is possible to happen on
> > emulators
> > (which emulate floppy operations extremely fast) as well as real
> > hardware
> > (during reset).
> This doesn't look correct to me. The SeaBIOS C code always runs with
> interrupts disabled. The FRS_IRQ flag is only set from the
> handle_0e() hardware interrupt handler. So, I don't think there is a
> race condition with the clearing and checking of that flag.
Sorry, I didn't know that the SeaBIOS C code runs with interrupts
disabled. I assumed they were enabled. Please disregard this patch.
No problem. Be aware that yield() or anything that calls yield()
(such as sleep() ) will allow irqs to occur. There's some
documentation on this in docs/Execution_and_code_flow.md.