auto generating task by Suryansh-Dey · Pull Request #441 · luainkernel/lunatik

Suryansh-Dey · 2026-02-21T05:07:43Z

Closes #435
Here is phase 1 of the PR

Simplify configure.lua (remove stat, signal, eth, supplement logic)
Restore lualinux_stat in lib/lualinux.c
Update documentation for task access

Now make configure auto generates hexadecimal constants for both task and stat modules.

Phase 2: Move stat constants
Phase 3: Move OR'ed flags (composites)
Ready for the feedback.

Suryansh-Dey · 2026-02-22T06:01:07Z

I did run the relevant tests and nothing seem to break:

~/dev/GSOC/lunatik task-phased-v2 [!]
❯ sudo lunatik reload
  sudo lunatik status
[sudo] password for suryansh:
lunatik is loaded
luadevice is loaded
lualinux is loaded
luanotifier is loaded
luasocket is loaded
luarcu is loaded
luathread is loaded
luafib is loaded
luadata is loaded
luaprobe is loaded
luasyscall is loaded
luaxdp is loaded
luafifo is loaded
luaxtable is loaded
luanetfilter is loaded
luacompletion is loaded
luacrypto_shash is loaded
luacrypto_skcipher is loaded
luacrypto_aead is loaded
luacrypto_rng is loaded
luacrypto_comp is loaded
luacpu is loaded
luahid is loaded
luasignal is loaded
luabyteorder is loaded
lunatik_run is loaded

~/dev/GSOC/lunatik task-phased-v2 [!]
❯ echo 'local task = require("linux.task"); print(task.INTERRUPTIBLE, task.UNINTERRUPTIBLE, task.KILLABLE, task.IDLE); return "task OK"' | sudo lunatik
  sudo dmesg | tail -5
Lunatik 4.1  Copyright (C) 2023-2026 Ring Zero Desenvolvimento de Software LTDA.
> task OK
> [  610.149922] 2
[  610.149924]
[  610.149925] 258
[  610.149927]
[  610.149928] 1026

~/dev/GSOC/lunatik task-phased-v2 [!]
❯ echo 'local linux = require("linux"); local task = require("linux.task"); linux.schedule(100, task.INTERRUPTIBLE); return "schedule OK"' | sudo lunatik

Lunatik 4.1  Copyright (C) 2023-2026 Ring Zero Desenvolvimento de Software LTDA.
> schedule OK
> ⏎

~/dev/GSOC/lunatik task-phased-v2 [!]
❯ echo 'local linux = require("linux"); print(linux.stat.IRUGO); return "stat OK"' | sudo lunatik
  sudo dmesg | tail -1
Lunatik 4.1  Copyright (C) 2023-2026 Ring Zero Desenvolvimento de Software LTDA.
> stat OK
> [  710.751080] 292

Got 292 as expected.

What should I do with examples/shared.lua and example/echod/daemon.lua? Remove the line to get rid of linter error?

lneto · 2026-02-22T12:49:43Z

I did run the relevant tests and nothing seem to break:

you did break the examples, right? we shouldn't break anything.

What should I do with examples/shared.lua and example/echod/daemon.lua? Remove the line to get rid of linter error?

you should adjust the examples accordingly.

Suryansh-Dey

I have moved

if tonumber(value) then return value end

to

	local n = tonumber(value)
	if n then return string.format("0x%08x", n) end

To avoid mixing octal and hex. Or should I keep it octal?
(Git diff seem to glitch a bit there so better you look at the new code instead of looking at the diff)

lneto · 2026-02-26T11:28:22Z

@Suryansh-Dey are you still working on this? please, let me know.

Suryansh-Dey · 2026-02-26T11:36:17Z

I have exams tomorrow and day after tomorrow, I will get back to it on 28th feb.

lneto · 2026-02-26T11:39:13Z

I have exams tomorrow and day after tomorrow, I will get back to it on 28th feb.

good luck! no worries, just wanted to check if we should keep this PR opened.. thanks!

lneto · 2026-03-01T18:41:43Z

-local function resolve_value(value, constants, prefix, module_name, seen)
-	if tonumber(value) then return value end
+local function resolve_value(value, constants, seen)
+	if tonumber(value) then return tonumber(value) end


Because write_constants now uses string.format('0x%08x', resolved_value) which requires an actual number, not a string. The original returned the raw string "0x00000001" and wrote it with %s.
We need it for:

Format it consistently as 0x%08x

Use the bitwise OR operator result | resolved for composite constants like TASK_KILLABLE

it seems odd to me to convert it twice.. perhaps you should use tonumber on the caller and if type is number here, return it.. not sure.. but you should clean up your logics.. it seems very convoluted to me..

lneto · 2026-03-01T18:43:26Z

@sneaky-potato can you please review it? It's your baby ;-)

sneaky-potato · 2026-03-06T06:35:55Z

+				is_or_expr = false
+				break
+			end
+			local resolved = resolve_value(constants[token] or token, prefix, constants, seen)


Currently, if seen is nil at the top level (first call), each component initializes its own independent seen in the calls of resolve_value. For it to work correctly, seen must be defined at the top of this function I think.

Yes, I have added some safety checks without compromising performance.

sneaky-potato · 2026-03-06T13:10:56Z

+	if tonumber(value) then return tonumber(value) end
 	seen = seen or {}
-	-- recursion check: avoid cyclic defines
-	if seen[value] then return nil end


Please don't move this. Keep this block above the main loop only, it will allow the method to return early if the macro has been seen.

remove [%w_]+ regex from resolve_value Updating makefile to provide arch Update Makefile Co-authored-by: Ashwani Kumar Kamal <75236490+sneaky-potato@users.noreply.github.com>

Suryansh-Dey · 2026-03-01T21:20:08Z

-local function resolve_value(value, constants, prefix, module_name, seen)
-	if tonumber(value) then return value end
+local function resolve_value(value, constants, seen)
+	if tonumber(value) then return tonumber(value) end


Because write_constants now uses string.format('0x%08x', resolved_value) which requires an actual number, not a string. The original returned the raw string "0x00000001" and wrote it with %s.
We need it for:

Format it consistently as 0x%08x

Use the bitwise OR operator result | resolved for composite constants like TASK_KILLABLE

Suryansh-Dey · 2026-03-06T07:43:56Z

+				is_or_expr = false
+				break
+			end
+			local resolved = resolve_value(constants[token] or token, prefix, constants, seen)


Yes, I have added some safety checks without compromising performance.

Suryansh-Dey · 2026-03-06T13:55:16Z

+	if tonumber(value) then return tonumber(value) end
 	seen = seen or {}
-	-- recursion check: avoid cyclic defines
-	if seen[value] then return nil end


Suryansh-Dey · 2026-03-08T10:24:45Z

 		module_name = "eth",
 	},
 	{
-		header = "uapi/linux/stat.h",


Switching to linux/stat.h ensures the CPP sees both the UAPI base constants and the internal compound ones like S_IRUGO = (S_IRUSR|S_IRGRP|S_IROTH), which our resolve_value function then resolves into their final numeric values.

Suryansh-Dey · 2026-03-09T02:57:32Z

All the tests are passing now. If the PR is ready to merge and no changes needed I can squash the commits. Please let me know

sneaky-potato · 2026-03-10T17:01:58Z

+local BUILD = INCLUDE:gsub("/include$", "")
+local CPP = string.format("%s -E -dM -I%s -I%s/include/generated -I%s/arch/%s/include -I%s/arch/%s/include/generated -include linux/kconfig.h",


I feel we should accept BUILD as a command line argument instead of INCLUDE to the script.
INCLUDE would then become BUILD .. "/include"
This way we can construct the other paths as well instead of relying on string manipulation.

If you notice in Makefile we already have a variable ${MODULES_BUILD_PATH}
What do you think?

as you wish

as you wish

don't you agree with the suggestions? if so, please make your point..

No, I do agree. I also think it is cleaner.

sneaky-potato · 2026-03-11T09:19:28Z

+    if clean:find("|") then
+        local result = 0
+        for token in clean:gmatch("[^%s|]+") do
+            local resolved = resolve_value(token, prefix, constants, seen)


Here we are not setting seen[value] = true right before this resolve_value call. This might cause problems right?

Actually, this won't cause any problems.
The value at this specific point is the literal OR expression string itself. We only need to track the actual macro names in the seen table to detect cycles. Since each individual token inside the OR expression (like TASK_A and TASK_B) gets passed back into resolve_value recursively, they will hit the seen[value] = true block at the bottom when they are resolved as macro references. If an OR expression contains a macro that forms a cycle, it will still be caught perfectly.
Please correct me If I missed something

sneaky-potato · 2026-03-11T09:20:49Z

@@ -12,6 +12,8 @@ MODULES_BUILD_PATH ?= ${BTF_INSTALL_PATH}
 MODULES_INSTALL_PATH := ${MODULES_RELEASE_PATH}/kernel
 SCRIPTS_INSTALL_PATH := ${MODULES_PATH}/lua
 INCLUDE_PATH := ${MODULES_BUILD_PATH}/include


This is not used anywhere now, we can safely remove this

MODULES_INSTALL_PATH := ${MODULES_RELEASE_PATH}/kernel SCRIPTS_INSTALL_PATH := ${MODULES_PATH}/lua

They get used later in the makefile below, so we can't remove them.
But yes I should have removed line 14. My bad.

sneaky-potato · 2026-03-11T09:32:17Z

 SCRIPTS_INSTALL_PATH := ${MODULES_PATH}/lua
 INCLUDE_PATH := ${MODULES_BUILD_PATH}/include
+KERNEL_ARCH ?= ${shell uname -m}
+ARCH ?= $(subst x86_64,x86,$(subst aarch64,arm64,$(KERNEL_ARCH)))


I am having mixed feelings here. We already do similar logic to find architecture here in KBuild. Maybe export this as environment variable so that both the Kbuild and configure.lua simply "consume" instead of recomputing it.
@lneto what do you think?

We should never reinvent the wheel.. BTW, we should care about cross compiling; particularly, this should work with OpenWRT feed as well..

What should I do then?

I think @sneaky-potato's idea is way to do it, just make sure it works on OpenWRT feed. It must be tested there as well.

Could you please test if works on OpenWRT feed. I am not sure of that thing, how it works.

I can help if you would like to do it, but not do it for you. For merging this PR, we need to make sure we don't break it. For building it you can just use our OpenWRT feed pointing to your commit. You probably will need to adjust the makefile there.

Co-authored-by: Ashwani Kumar Kamal <75236490+sneaky-potato@users.noreply.github.com>

Suryansh-Dey mentioned this pull request Feb 21, 2026

Moving to task #440

Closed

lneto reviewed Feb 21, 2026

View reviewed changes

Comment thread lib/lualinux.c Outdated

lneto reviewed Feb 21, 2026

View reviewed changes

Comment thread configure.lua Outdated

lneto requested review from carloslack and sneaky-potato February 21, 2026 11:31

Suryansh-Dey force-pushed the task-phased-v2 branch from 2a921b7 to 359fe69 Compare February 21, 2026 11:59

lneto reviewed Feb 21, 2026

View reviewed changes

Comment thread lib/lualinux.c

sneaky-potato reviewed Feb 22, 2026

View reviewed changes

Comment thread configure.lua Outdated

Suryansh-Dey force-pushed the task-phased-v2 branch 2 times, most recently from 34d4bb1 to fcb04eb Compare February 22, 2026 12:38

Suryansh-Dey commented Feb 22, 2026

View reviewed changes

Comment thread configure.lua Outdated

Comment thread lib/lualinux.c

Suryansh-Dey force-pushed the task-phased-v2 branch 2 times, most recently from 5d75ba8 to 64c7234 Compare February 22, 2026 14:29

Suryansh-Dey force-pushed the task-phased-v2 branch 3 times, most recently from 804d1c7 to 55e8947 Compare March 1, 2026 17:48

lneto reviewed Mar 1, 2026

View reviewed changes

Comment thread configure.lua Outdated

lneto reviewed Mar 1, 2026

View reviewed changes

Comment thread configure.lua Outdated

Suryansh-Dey force-pushed the task-phased-v2 branch from 55e8947 to 244eeb3 Compare March 2, 2026 03:21

sneaky-potato reviewed Mar 6, 2026

View reviewed changes

Comment thread configure.lua Outdated

Suryansh-Dey force-pushed the task-phased-v2 branch 3 times, most recently from 6cec57b to 5f7de2f Compare March 6, 2026 08:18

Suryansh-Dey requested a review from sneaky-potato March 6, 2026 08:22

Suryansh-Dey force-pushed the task-phased-v2 branch from 5f7de2f to 482f434 Compare March 6, 2026 08:29

sneaky-potato reviewed Mar 6, 2026

View reviewed changes

Suryansh-Dey force-pushed the task-phased-v2 branch from 482f434 to 972b585 Compare March 6, 2026 13:51

Suryansh-Dey requested a review from sneaky-potato March 6, 2026 13:55

sneaky-potato reviewed Mar 6, 2026

View reviewed changes

Comment thread Makefile Outdated

Suryansh-Dey force-pushed the task-phased-v2 branch from 311bc38 to 4454abb Compare March 6, 2026 17:46

Suryansh-Dey requested a review from sneaky-potato March 6, 2026 17:55

using cc preprocessor

501ad9f

remove [%w_]+ regex from resolve_value Updating makefile to provide arch Update Makefile Co-authored-by: Ashwani Kumar Kamal <75236490+sneaky-potato@users.noreply.github.com>

Suryansh-Dey force-pushed the task-phased-v2 branch 2 times, most recently from 4a1a811 to 57c4715 Compare March 8, 2026 10:17

Removing hard coded S_* to auto-generate it

914fc2b

Suryansh-Dey force-pushed the task-phased-v2 branch from 57c4715 to 914fc2b Compare March 8, 2026 10:18

Suryansh-Dey commented Mar 8, 2026

View reviewed changes

sneaky-potato reviewed Mar 9, 2026

View reviewed changes

Comment thread configure.lua Outdated

sneaky-potato reviewed Mar 10, 2026

View reviewed changes

Suryansh-Dey force-pushed the task-phased-v2 branch from eb3f494 to a62618f Compare March 11, 2026 08:58

sneaky-potato reviewed Mar 11, 2026

View reviewed changes

Committing to pull and test

bffdd84

Co-authored-by: Ashwani Kumar Kamal <75236490+sneaky-potato@users.noreply.github.com>

Suryansh-Dey force-pushed the task-phased-v2 branch from a62618f to bffdd84 Compare March 12, 2026 14:27

Making arch env var

48abad8

lneto closed this Apr 18, 2026

		local BUILD = INCLUDE:gsub("/include$", "")
		local CPP = string.format("%s -E -dM -I%s -I%s/include/generated -I%s/arch/%s/include -I%s/arch/%s/include/generated -include linux/kconfig.h",

Conversation

Suryansh-Dey commented Feb 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Suryansh-Dey commented Feb 22, 2026

Uh oh!

Uh oh!

lneto commented Feb 22, 2026

Uh oh!

Suryansh-Dey left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

lneto commented Feb 26, 2026

Uh oh!

Suryansh-Dey commented Feb 26, 2026

Uh oh!

lneto commented Feb 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

lneto commented Mar 1, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

sneaky-potato Mar 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Suryansh-Dey commented Mar 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

sneaky-potato Mar 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Suryansh-Dey commented Feb 21, 2026 •

edited

Loading

Suryansh-Dey left a comment •

edited

Loading

sneaky-potato Mar 6, 2026 •

edited

Loading

Suryansh-Dey commented Mar 9, 2026 •

edited

Loading

sneaky-potato Mar 10, 2026 •

edited

Loading

Suryansh-Dey Mar 12, 2026 •

edited

Loading