Indan Zupancic

For MCS, the server should be passive too.

Moving the yield won't help this particular issue though, as the time is too long, not too short. Do we even now for sure that the clock used for scheduling...

The timestamp is added by the logging system and includes the loop time, so it just shows a ridiculously slow (buggy?) UART driver (2 ms per character). The quality of...

How long before the first ARM system actually has more than 128 TB of SDRAM memory and this becomes a practical problem? Virtual address space size seems to be 52-bits...

The main information that is extracted from the device trees are the memory regions. But for devices that's fixed per SoC, usually it are only a few fixed large regions....

For level triggered interrupts you need to tell the device you handled them before ACKing. If the kernel ACKs them automatically, the same interrupt will just trigger immediately. So this...

> > However, edge-triggered interrupts are not performance critical, they per definition don't happen often (they would be level triggered otherwise). > > This seems a surprising claim to me?...

> When this ack is unconditional and happens before handling the interrupt, its redundant for the kernel to require a separate call for the ack. Agreed it's just a hassle...

> That's why I'd like to see implicit ACKin on waiting on the next IRQ. But this introduces a race condition between the last device check and the implicit ACK....

> @Indanz, After thinking a bit more, I realize that acking earlier than the receiving thread becoming scheduled would introduce an unbounded denial-of-service ability Well, that is always the problem...