capsules: Convert analog comparator to use grant #2521

bradjc · 2021-04-08T21:01:44Z

Pull Request Overview

As part of #2462 this switches the analog comparator capsule to use a grant.

Testing Strategy

travis

TODO or Help Wanted

n/a

Documentation Updated

Updated the relevant files in /docs, or no updates are required.

Formatting

Ran make prepush.

capsules/src/analog_comparator.rs

bradjc · 2021-04-08T21:13:51Z

Good point. I think that is worth an extra check.

phil-levis · 2021-04-10T00:00:19Z

bors r+

bors · 2021-04-10T00:00:33Z

Merge conflict.

lschuermann · 2021-04-12T19:45:18Z

capsules/src/analog_comparator.rs

+        if match_or_empty_or_nonexistant {
+            self.current_process.set(appid);
+        } else {
+            return CommandReturn::failure(ErrorCode::NOMEM);


Intuitively I would've expected another ErrorCode, something like BUSY? NOMEM kind of fits here I guess?

Historically that is the error that has been returned because there was no memory available to store the new process's request when grants were not being used.

Maybe we should use ErrorCode::RESERVE? I'm not that worried either way as I don't think userspace will actually use these and they should be virtualized anyway.

The explanation behind NOMEM makes sense, I'm happy to use that. I went through the other error codes and I agree, no other code seems to fit well.

I'll update #2518 for consistency's sake.

lschuermann · 2021-04-12T21:11:30Z

I've just run a minimal example of how this PR checks whether a nonvirtualized driver can be reserved by a process, compared to the implementation in #2518. This was only for curiosity's sake, and to see whether entering the Grant would have significant impact. These numbers are obtained on a rv32imc build with rustc 1.52.0-nightly (1705a7d64 2021-03-18) and this minimal example. I measured the system call end-to-end instruction count, that is # of instructions executed from the trap handler, until the mret returning to the process.

Using this approach, on the first command syscall (cmd_num = 1) I measure 903, on the second 1387, and on the third and subsequent calls 1034 instructions.
Using the approach of Lsm303dlhc: Enforce use by a single process, move callback to Grant #2518, the first command syscall takes 890, and subsequent 887.

This is probably fine for non-virtualized drivers (given they aren't Tock's idiomatic way to write drivers). It does however seem to tell us that entering a Grant has a cost which one might want to look into optimizing, along with system call routing and switching to a process. Here is a more thorough investigation of the instructions spent on syscall routing.

Used in more places than I thought!

ppannuto · 2021-04-14T20:31:53Z

bors r+

bors · 2021-04-14T20:43:48Z

Build succeeded:

As part of tock#2462, this migrates the spi_peripheral non-virtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in tock#2521. Signed-off-by: Leon Schuermann <[email protected]>

As part of tock#2462, this migrates the sdcard non-virtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in tock#2521. Signed-off-by: Leon Schuermann <[email protected]>

As part of tock#2462, this migrates the spi_controller nonvirtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in tock#2521. Signed-off-by: Leon Schuermann <[email protected]>

2559: capsules/spi_peripheral: migrate non-virtualized driver to Grants r=bradjc a=lschuermann ### Pull Request Overview As part of #2462, this migrates the spi_peripheral non-virtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in #2521. Signed-off-by: Leon Schuermann <[email protected]> ### Testing Strategy This pull request was tested by compiling. ### TODO or Help Wanted N/A ### Documentation Updated - [x] ~Updated the relevant files in `/docs`,~ or no updates are required. ### Formatting - [x] Ran `make prepush`. 2560: capsules/sdcard: migrate non-virtualized userspace driver to Grants r=bradjc a=lschuermann ### Pull Request Overview As part of #2462, this migrates the sdcard non-virtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in #2521. Signed-off-by: Leon Schuermann <[email protected]> ### Testing Strategy This pull request was tested by compiling. ### TODO or Help Wanted N/A ### Documentation Updated - [x] ~Updated the relevant files in `/docs`,~ or no updates are required. ### Formatting - [x] Ran `make prepush`. Co-authored-by: Leon Schuermann <[email protected]>

2547: capsules/i2c_master_slave_driver: use grant, enforce single process r=bradjc a=hudson-ayers ### Pull Request Overview This pull request enforces that only a single process may (implicitly) reserve the nonvirtualized userspace driver. Also, the driver now store callbacks and appslices in grant regions. This is one of the remaining capsules blocking #2462 . One observation: repeatedly handling the possibility of a grant entrance (within a single function) failing is a pain. But I do not think it is possible for a grant entrance to fail after the first entrance -- the grant is already allocated at that point, and the process is present. So I omitted the error handling in `command()` after the first grant entrance. ### Testing Strategy This pull request was tested by compiling. ### TODO or Help Wanted N/A ### Documentation Updated - [x] No updates are required. ### Formatting - [x] Ran `make prepush`. 2558: capsules/spi_controller: migrate non-virtualized driver to Grants r=bradjc a=lschuermann ### Pull Request Overview As part of #2462, this migrates the spi_controller nonvirtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in #2521. Signed-off-by: Leon Schuermann <[email protected]> ### Testing Strategy This pull request was tested by compiling. ### TODO or Help Wanted N/A ### Documentation Updated - [x] ~Updated the relevant files in `/docs`, or~ no updates are required. ### Formatting - [x] Ran `make prepush`. Co-authored-by: Hudson Ayers <[email protected]> Co-authored-by: Leon Schuermann <[email protected]>

2572: capsules/l3gd20: migrating non-virtualized driver to Grants r=hudson-ayers a=phil-levis ### Pull Request Overview As part of #2462, this migrates the l3gd20 nonvirtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in #2521. ### Testing Strategy This pull request was tested by compiling. ### TODO or Help Wanted N/A ### Documentation Updated - [x] Updated the relevant files in `/docs`, or no updates are required. ### Formatting - [x] Ran `make prepush`. Co-authored-by: Philip Levis <[email protected]>

2575: capsules/ltc294x-grant: migrating non-virtualized driver to Grants r=hudson-ayers a=phil-levis ### Pull Request Overview As part of #2462, this migrates the ltc294x nonvirtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in #2521. ### Testing Strategy This pull request was tested by compiling. ### TODO or Help Wanted N/A ### Documentation Updated - [x] Updated the relevant files in `/docs`, or no updates are required. ### Formatting - [x] Ran `make prepush`. Co-authored-by: Philip Levis <[email protected]>

2574: capsules/lsm303agr: migrating non-virtualized driver to Grants r=hudson-ayers a=phil-levis ### Pull Request Overview As part of #2462, this migrates the lsm303agr nonvirtualized userspace driver to keep process state in a Grant region. It adopts the mechanism to reserve a driver as introduced in #2521. ### Testing Strategy This pull request was tested by compiling. ### TODO or Help Wanted N/A ### Documentation Updated - [x] Updated the relevant files in `/docs`, or no updates are required. ### Formatting - [x] Ran `make prepush`. Co-authored-by: Philip Levis <[email protected]>

hudson-ayers previously approved these changes Apr 8, 2021

View reviewed changes

capsules/src/analog_comparator.rs Show resolved Hide resolved

bradjc dismissed hudson-ayers’s stale review via 0daebbd April 8, 2021 21:12

bradjc force-pushed the grant-analog-comparator branch from 4900288 to 0daebbd Compare April 8, 2021 21:12

hudson-ayers previously approved these changes Apr 8, 2021

View reviewed changes

bradjc dismissed hudson-ayers’s stale review via 3d02227 April 9, 2021 15:15

hudson-ayers previously approved these changes Apr 9, 2021

View reviewed changes

alistair23 previously approved these changes Apr 11, 2021

View reviewed changes

lschuermann reviewed Apr 12, 2021

View reviewed changes

lschuermann mentioned this pull request Apr 12, 2021

Lsm303dlhc: Enforce use by a single process, move callback to Grant #2518

Merged

2 tasks

hudson-ayers mentioned this pull request Apr 13, 2021

tock-cells/OptionalCell: add get-method for retrieving inner Option #2531

Merged

2 tasks

lschuermann previously approved these changes Apr 14, 2021

View reviewed changes

bradjc added 2 commits April 14, 2021 13:07

capsules: analog comparator use grant

e9621aa

boards: update analog comparator

c353010

Used in more places than I thought!

bradjc dismissed stale reviews from lschuermann, alistair23, and hudson-ayers via c353010 April 14, 2021 17:08

bradjc force-pushed the grant-analog-comparator branch from 3d02227 to c353010 Compare April 14, 2021 17:08

hudson-ayers approved these changes Apr 14, 2021

View reviewed changes

lschuermann approved these changes Apr 14, 2021

View reviewed changes

bors bot merged commit 4a29f38 into master Apr 14, 2021

bors bot deleted the grant-analog-comparator branch April 14, 2021 20:43

This was referenced Apr 16, 2021

Tock 2.0: implement Callback swapping restrictions (v3) #2462

Closed

capsules/i2c_master_slave_driver: use grant, enforce single process #2547

Merged

lschuermann mentioned this pull request Apr 26, 2021

capsules/spi_controller: migrate non-virtualized driver to Grants #2558

Merged

2 tasks

lschuermann mentioned this pull request Apr 26, 2021

capsules/spi_peripheral: migrate non-virtualized driver to Grants #2559

Merged

2 tasks

lschuermann mentioned this pull request Apr 26, 2021

capsules/sdcard: migrate non-virtualized userspace driver to Grants #2560

Merged

2 tasks

This was referenced May 4, 2021

capsules/l3gd20: migrating non-virtualized driver to Grants #2572

Merged

capsules/lsm303agr: migrating non-virtualized driver to Grants #2574

Merged

phil-levis mentioned this pull request May 5, 2021

capsules/ltc294x-grant: migrating non-virtualized driver to Grants #2575

Merged

2 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

capsules: Convert analog comparator to use grant #2521

capsules: Convert analog comparator to use grant #2521

Uh oh!

bradjc commented Apr 8, 2021

Uh oh!

Uh oh!

bradjc commented Apr 8, 2021

Uh oh!

phil-levis commented Apr 10, 2021

Uh oh!

bors bot commented Apr 10, 2021

Uh oh!

lschuermann Apr 12, 2021

Uh oh!

bradjc Apr 14, 2021

Uh oh!

lschuermann Apr 14, 2021

Uh oh!

lschuermann commented Apr 12, 2021 •

edited

Loading

Uh oh!

ppannuto commented Apr 14, 2021

Uh oh!

bors bot commented Apr 14, 2021

Uh oh!

Uh oh!

Uh oh!

capsules: Convert analog comparator to use grant #2521

capsules: Convert analog comparator to use grant #2521

Uh oh!

Conversation

bradjc commented Apr 8, 2021

Pull Request Overview

Testing Strategy

TODO or Help Wanted

Documentation Updated

Formatting

Uh oh!

Uh oh!

bradjc commented Apr 8, 2021

Uh oh!

phil-levis commented Apr 10, 2021

Uh oh!

bors bot commented Apr 10, 2021

Uh oh!

lschuermann Apr 12, 2021

Choose a reason for hiding this comment

Uh oh!

bradjc Apr 14, 2021

Choose a reason for hiding this comment

Uh oh!

lschuermann Apr 14, 2021

Choose a reason for hiding this comment

Uh oh!

lschuermann commented Apr 12, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ppannuto commented Apr 14, 2021

Uh oh!

bors bot commented Apr 14, 2021

Uh oh!

Uh oh!

lschuermann commented Apr 12, 2021 •

edited

Loading