rust : Add RROS_SPINLOCK flag and adjust spinlock initialization for new constructor #66

Caspian443 · 2024-11-08T02:49:48Z

Add another config RROS_SPINLOCK
Modify some CONFIG_RROS to CONFIG_RROS_SPINLOCK
Adjust the spinlock initialization in rust/kernel/memory_rros.rs to accommodate the new constructor

shannmu

BTW, there is a CI failure. Please make each commit pass compile.

shannmu · 2024-11-08T08:55:31Z

rust/kernel/memory_rros.rs

+        let mut spinlock_init =  unsafe { SpinLock::new(1, lock_name, lock_key) } ;
+        // let pinned = unsafe { Pin::new_unchecked(&mut spinlock) };
+        // spinlock_init!(pinned, "spinlock");
+        let mut spinlock= Box::pin_init(spinlock_init).unwrap();


There is a macro named new_spinlock. I think you could try it.

I'm so careless,I'll try it.

Caspian443 · 2024-11-08T09:23:54Z

BTW, there is a CI failure. Please make each commit pass compile.
Thank you for the feedback. I’ll ensure that each commit passes compilation and address the CI failure.

…ble.rs

…l.rs

…observable.rs

Caspian443 · 2024-11-10T15:28:58Z

kernel/rros/idle.rs

@@ -94,7 +94,7 @@ fn __rros_set_idle_schedparam(
    // let mut thread_lock = thread_unwrap.lock();
    let p_unwrap = p.unwrap();
    thread_unwrap.lock().state &= !T_WEAK;
-    let prio = unsafe { (*p_unwrap.locked_data().get()).idle.prio };
+    let prio = unsafe { (*p_unwrap.lock().deref()).idle.prio };


cc @shannmu

I don't know why there is a locked_data rather than a lock. I think the lock is better here.

Indeed, the version now has only lock.

Caspian443 · 2024-11-10T15:29:55Z

kernel/rros/poll.rs

+        // spinlock_init!(
+        //     unsafe { Pin::new_unchecked(&mut self.watchpoints) },
+        //     "RrosPollHead"
+        // );
    }
 }


cc @shannmu

Good for me.

Caspian443 · 2024-11-10T15:30:53Z

kernel/rros/net/device.rs

+// init_static_sync! {
+//     static ACTIVE_PORT_LIST : SpinLock<ListThreadSafeWrapper> = ListThreadSafeWrapper(List::new());
+// }
+static ACTIVE_PORT_LIST : Pin<Box<SpinLock<ListThreadSafeWrapper>>> = Box::pin_init(new_spinlock!(ListThreadSafeWrapper(List::new()))).unwrap();



cc @shannmu

It is good for me for now. Box::pin_init returns a Result and using unwrap here is a temporary solution. I believe we need to address this problem later.

Caspian443 · 2024-11-25T13:37:50Z

I have aligned the new class_create function, and class_new can successfully return. However, the program still terminates at this position（waiting for a long time）.

The error is like this:

Caspian443 · 2024-11-25T13:47:16Z

I would like to show you the process of debugging with gdb in this video
https://github.com/user-attachments/assets/af4d3cd0-4a10-4660-983a-a86f5ece5ee1

shannmu · 2024-11-25T13:49:36Z

However, the program still terminates at this position（waiting for a long time）.

I think this ret_from_fork means it is in a thread stask. And the do_populate_rootfs is to load the built-in initramfs. So init of RROS is good and there may be an error for your filesystem

Fix: Fix the compile error for RROS

Caspian443 · 2024-12-06T13:25:20Z

kernel/rros/clock.rs

@@ -1218,7 +1218,7 @@ fn init_clock(clock: *mut RrosClock, master: *mut RrosClock) -> Result<usize> {
        clocklist_lock_init();
        CLOCKLIST_LOCK.get().unwrap().lock();
        CLOCK_LIST.add_head(clock);
-        CLOCKLIST_LOCK.get().unwrap().unlock();
+        CLOCKLIST_LOCK.get().unwrap().unlock(&());


The reason I added this parameter is that GuardState is an empty tuple (), indicating that no additional state needs to be maintained between locking and unlocking.

Caspian443 · 2024-12-06T13:27:46Z

rust/kernel/sync/lock.rs

        unsafe {
-            rust_helper_hard_spin_unlock(self.lock().lock as *const Lock<_, _>  as *mut bindings::raw_spinlock)


I eliminated lock() by using the unlock of the Backend.

…b folio A kernel crash was observed when migrating hugetlb folio: BUG: kernel NULL pointer dereference, address: 0000000000000008 PGD 0 P4D 0 Oops: Oops: 0002 [BUPT-OS#1] PREEMPT SMP NOPTI CPU: 0 PID: 3435 Comm: bash Not tainted 6.10.0-rc6-00450-g8578ca01f21f BUPT-OS#66 RIP: 0010:__folio_undo_large_rmappable+0x70/0xb0 RSP: 0018:ffffb165c98a7b38 EFLAGS: 00000097 RAX: fffffbbc44528090 RBX: 0000000000000000 RCX: 0000000000000000 RDX: ffffa30e000a2800 RSI: 0000000000000246 RDI: ffffa3153ffffcc0 RBP: fffffbbc44528000 R08: 0000000000002371 R09: ffffffffbe4e5868 R10: 0000000000000001 R11: 0000000000000001 R12: ffffa3153ffffcc0 R13: fffffbbc44468000 R14: 0000000000000001 R15: 0000000000000001 FS: 00007f5b3a716740(0000) GS:ffffa3151fc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000008 CR3: 000000010959a000 CR4: 00000000000006f0 Call Trace: <TASK> __folio_migrate_mapping+0x59e/0x950 __migrate_folio.constprop.0+0x5f/0x120 move_to_new_folio+0xfd/0x250 migrate_pages+0x383/0xd70 soft_offline_page+0x2ab/0x7f0 soft_offline_page_store+0x52/0x90 kernfs_fop_write_iter+0x12c/0x1d0 vfs_write+0x380/0x540 ksys_write+0x64/0xe0 do_syscall_64+0xb9/0x1d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7f5b3a514887 RSP: 002b:00007ffe138fce68 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007f5b3a514887 RDX: 000000000000000c RSI: 0000556ab809ee10 RDI: 0000000000000001 RBP: 0000556ab809ee10 R08: 00007f5b3a5d1460 R09: 000000007fffffff R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c R13: 00007f5b3a61b780 R14: 00007f5b3a617600 R15: 00007f5b3a616a00 It's because hugetlb folio is passed to __folio_undo_large_rmappable() unexpectedly. large_rmappable flag is imperceptibly set to hugetlb folio since commit f6a8dd9 ("hugetlb: convert alloc_buddy_hugetlb_folio to use a folio"). Then commit be9581e ("mm: fix crashes from deferred split racing folio migration") makes folio_migrate_mapping() call folio_undo_large_rmappable() triggering the bug. Fix this issue by clearing large_rmappable flag for hugetlb folios. They don't need that flag set anyway. Link: https://lkml.kernel.org/r/[email protected] Fixes: f6a8dd9 ("hugetlb: convert alloc_buddy_hugetlb_folio to use a folio") Fixes: be9581e ("mm: fix crashes from deferred split racing folio migration") Signed-off-by: Miaohe Lin <[email protected]> Cc: Hugh Dickins <[email protected]> Cc: Matthew Wilcox (Oracle) <[email protected]> Cc: Muchun Song <[email protected]> Cc: <[email protected]> Signed-off-by: Andrew Morton <[email protected]>

Interrupts on/off mismatch when locking the recycling queue. At this chance annotate some routines wrt the legit call context. [ 39.474194] ================================ [ 39.474195] WARNING: inconsistent lock state [ 39.474197] 6.1.54+ BUPT-OS#66 Not tainted [ 39.474200] -------------------------------- [ 39.474201] inconsistent {HARDIRQ-ON-W} -> {IN-HARDIRQ-W} usage. [ 39.474203] eth0.rx/370 [HC1[0]:SC0[0]:HE0:SE1] takes: [ 39.474207] ffffffff8276b8d8 (recycling_lock){?.+.}-{0:0}, at: free_skb_inband+0x5/0xa0 [ 39.474222] {HARDIRQ-ON-W} state was registered at: [ 39.474223] __lock_acquire+0x363/0x9a0 [ 39.474229] lock_acquire+0xbe/0x2a0 [ 39.474233] free_skb_inband+0x2a/0xa0 [ 39.474236] evl_net_free_skb+0x11/0x90 [ 39.474239] evl_net_do_rx+0x1b7/0x280 [ 39.474242] kthread_trampoline+0x1c7/0x2d0 [ 39.474246] kthread+0xf5/0x120 [ 39.474251] ret_from_fork+0x22/0x30 [ 39.474255] irq event stamp: 24 [ 39.474256] hardirqs last enabled at (23): [<ffffffff81b06125>] _raw_spin_unlock_irqrestore+0x65/0x80 [ 39.474262] hardirqs last disabled at (24): [<ffffffff81afb971>] __schedule+0x3a1/0x770 [ 39.474266] softirqs last enabled at (0): [<ffffffff810c1ad6>] copy_process+0x796/0x18c0 [ 39.474271] softirqs last disabled at (0): [<0000000000000000>] 0x0 [ 39.474274] [ 39.474274] other info that might help us debug this: [ 39.474275] Possible unsafe locking scenario: [ 39.474275] [ 39.474276] CPU0 [ 39.474276] ---- [ 39.474277] lock(recycling_lock); [ 39.474278] <Interrupt> [ 39.474279] lock(recycling_lock); [ 39.474280] [ 39.474280] *** DEADLOCK *** Signed-off-by: Philippe Gerum <[email protected]>

Caspian443 added 2 commits November 8, 2024 01:28

rust: rros: add 'CONFIG_RROS_SPINLOCK'

d897f9a

rust: rros: change some 'CONFIG_RROS' to 'CONFIG_RROS_SPINLOCK'

aa2df1d

Caspian443 changed the title ~~V6.6.y rros rebase~~ rust : Add RROS_SPINLOCK flag and adjust spinlock initialization for new constructor Nov 8, 2024

Caspian443 marked this pull request as draft November 8, 2024 02:57

shannmu reviewed Nov 8, 2024

View reviewed changes

change spinlock initialization in memory_rros.rs

3c18d74

Caspian443 force-pushed the v6.6.y-rros-rebase branch from b25fbeb to 3c18d74 Compare November 8, 2024 13:13

Caspian443 added 9 commits November 9, 2024 13:33

change spinlock initialization for RROS_CONTROL_FACTORY in control.rs

27d9d77

change spinlock initialization for RROS_OBSERVABLE_FACTORY in observa…

78397e2

…ble.rs

change spinlock initialization for RROS_POLL_FACTORY in poll.rs

ce619a1

change spinlock initialization for RrosPollGroup's waiter_list in pol…

9e1871f

…l.rs

change spinlock initialization for RrosPollHead's watchpoints in poll.rs

b9728a4

change spinlock initialization for RrosSubscriber's subscriptions in …

220446e

…observable.rs

change spinlock initialization for RROS_THREAD_FACTORY in thread.rs

d6369f7

Replace init_static_sync with the new spinlock initialization method.

37a4295

rust: rros: Refactor the method for accessing data within the lock

aaff261

Caspian443 commented Nov 10, 2024

View reviewed changes

Caspian443 added 6 commits November 23, 2024 13:13

change spinlock initialization

e68dfdb

reorganize statements for module imports

384acdf

wrap SpinLock with Pin and Box

47abfa2

add some necessary api

8e2e824

wrap global variable with Oncecell

fcfe82e

adjust class_create

4884575

rust: class: rm rust_helper_class_create

d6005dd

shannmu and others added 4 commits December 6, 2024 07:16

rust: time_types: mv time_types from uapi to kernel, fix compile error

b2921b7

Merge pull request #1 from shannmu/pr-66

63377a2

Fix: Fix the compile error for RROS

fix dead lock in spinlock's unlock

8154918

add a missing parameter.

85bac65

Caspian443 commented Dec 6, 2024

View reviewed changes

Caspian443 added 2 commits December 6, 2024 13:47

delete unlock

b27d418

impl unsafe operation for spinlock

ec59aa3

		unsafe {
		rust_helper_hard_spin_unlock(self.lock().lock as const Lock<_, _> as mut bindings::raw_spinlock)

rust : Add RROS_SPINLOCK flag and adjust spinlock initialization for new constructor #66

Are you sure you want to change the base?

rust : Add RROS_SPINLOCK flag and adjust spinlock initialization for new constructor #66

Uh oh!

Conversation

Caspian443 commented Nov 8, 2024

Uh oh!

shannmu left a comment

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!

Caspian443 commented Nov 8, 2024

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Caspian443 commented Nov 25, 2024

Uh oh!

Caspian443 commented Nov 25, 2024

Uh oh!

shannmu commented Nov 25, 2024

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!