Do allow buffers need to be static and what can the allow numbers be?

[potto216]

The capsule documentation talks about buffers needing a static lifetime.

But the console capsule doesn't state that as a requirement and the console code in libtock-rs/apis/console/src/lib.rs doesn't appear to require that. Is there any requirement that the lifetime is static?

My understanding of the allow numbers is that

1. They are independent per capsule.
2. In a capsule the read only numbers are an independent set from the read write numbers.

When I look at the documentation for the console capsule this is the rule they follow. Although they don't start at zero.

But when I look at the documentation for the udp capsule the read only and read write seem to be interleaved with one set of numbers.

What is the correct way to look at this?

[jrvanwhy]

The capsule documentation talks about buffers needing a static lifetime.

That documentation is about buffers within the kernel, not buffers shared between userspace processes and the kernel.

But the console capsule doesn't state that as a requirement and the console code in libtock-rs/apis/console/src/lib.rs doesn't appear to require that. Is there any requirement that the lifetime is static?

There is no requirement that Allow buffers be 'static.

My understanding of the allow numbers is that

1. They are independent per capsule.
2. In a capsule the read only numbers are an independent set from the read write numbers.

That understanding is correct.

When I look at the documentation for the console capsule this is the rule they follow. Although they don't start at zero.

But when I look at the documentation for the udp capsule the read only and read write seem to be interleaved with one set of numbers.

I believe this is a case of outdated documentation. In Tock 1.0 there was only one Allow system call, so read-only and read-write buffers had to have distinct buffer numbers. In Tock 2.0, they are different system calls, with independent numbering.

[hudson-ayers]

Yep, the UDP capsule was written before we split up buffers into RO / RW. The documentation is indeed out of date, but the numbers are in fact contiguous if you look at the code (https://github.com/tock/tock/blob/8655505563b27569dbf872c20c10144e517faa47/capsules/src/net/udp/driver.rs#L38-L51)

[potto216]

Thanks for the quick replies clearing up my confusion! I'll update the UDP documentation since I am working with that capsule.