From 1cc99630cfe4d9ef76c9b29bff4c0a8b49dab859 Mon Sep 17 00:00:00 2001 From: Mingcong Bai Date: Tue, 15 Oct 2024 20:32:21 +0800 Subject: [PATCH] AOSCOS: Revert "rcu: Fix rcu_barrier() VS post CPUHP_TEARDOWN_CPU invocation" When this change was introduced between v6.10.4 and v6.10.5, the Broadcom Tigon3 Ethernet interface (tg3) found on Apple MacBook Pro (15'', Mid 2010) would throw many rcu stall errors during boot up, causing peripherals such as the wireless card to misbehave. [ 24.153855] rcu: INFO: rcu_preempt detected expedited stalls on CPUs/tasks: { 2-.... } 21 jiffies s: 973 root: 0x4/. [ 24.166938] rcu: blocking rcu_node structures (internal RCU debug): [ 24.177800] Sending NMI from CPU 3 to CPUs 2: [ 24.183113] NMI backtrace for cpu 2 [ 24.183119] CPU: 2 PID: 1049 Comm: NetworkManager Not tainted 6.10.5-aosc-main #1 [ 24.183123] Hardware name: Apple Inc. MacBookPro6,2/Mac-F22586C8, BIOS MBP61.88Z.005D.B00.1804100943 04/10/18 [ 24.183125] RIP: 0010:__this_module+0x2d3d1/0x4f310 [tg3] [ 24.183135] Code: c3 cc cc cc cc 0f 1f 40 00 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 f3 0f 1e fa 0f 1f 44 00 00 89 f6 48 03 77 30 8b 06 <31> f6 31 ff c3 cc cc cc cc 66 0f 1f 44 00 00 90 90 90 90 90 90 90 [ 24.183138] RSP: 0018:ffffbf1a011d75e8 EFLAGS: 00000082 [ 24.183141] RAX: 0000000000000000 RBX: ffffa04ec78f8a00 RCX: 0000000000000000 [ 24.183143] RDX: 0000000000000000 RSI: ffffbf1a00fb007c RDI: ffffa04ec78f8a00 [ 24.183145] RBP: 0000000000000b50 R08: 0000000000000000 R09: 0000000000000000 [ 24.183147] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000216 [ 24.183148] R13: ffffbf1a011d7624 R14: ffffa04ec78f8a08 R15: ffffa04ec78f8b40 [ 24.183151] FS: 00007f4c524b2140(0000) GS:ffffa05007d00000(0000) knlGS:0000000000000000 [ 24.183153] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 24.183155] CR2: 00007f7025eae3e8 CR3: 00000001040f8000 CR4: 00000000000006f0 [ 24.183157] Call Trace: [ 24.183162] [ 24.183167] ? nmi_cpu_backtrace+0xbf/0x140 [ 24.183175] ? nmi_cpu_backtrace_handler+0x11/0x20 [ 24.183181] ? nmi_handle+0x61/0x160 [ 24.183186] ? default_do_nmi+0x42/0x110 [ 24.183191] ? exc_nmi+0x1bd/0x290 [ 24.183194] ? end_repeat_nmi+0xf/0x53 [ 24.183203] ? __this_module+0x2d3d1/0x4f310 [tg3] [ 24.183207] ? __this_module+0x2d3d1/0x4f310 [tg3] [ 24.183210] ? __this_module+0x2d3d1/0x4f310 [tg3] [ 24.183213] [ 24.183214] [ 24.183215] __this_module+0x31828/0x4f310 [tg3] [ 24.183218] ? __this_module+0x2d390/0x4f310 [tg3] [ 24.183221] __this_module+0x398e6/0x4f310 [tg3] [ 24.183225] __this_module+0x3baf8/0x4f310 [tg3] [ 24.183229] __this_module+0x4733f/0x4f310 [tg3] [ 24.183233] ? _raw_spin_unlock_irqrestore+0x25/0x70 [ 24.183237] ? __this_module+0x398e6/0x4f310 [tg3] [ 24.183241] __this_module+0x4b943/0x4f310 [tg3] [ 24.183244] ? delay_tsc+0x89/0xf0 [ 24.183249] ? preempt_count_sub+0x51/0x60 [ 24.183254] __this_module+0x4be4b/0x4f310 [tg3] [ 24.183258] __dev_open+0x103/0x1c0 [ 24.183265] __dev_change_flags+0x1bd/0x230 [ 24.183269] ? rtnl_getlink+0x362/0x400 [ 24.183276] dev_change_flags+0x26/0x70 [ 24.183280] do_setlink+0xe16/0x11f0 [ 24.183286] ? __nla_validate_parse+0x61/0xd40 [ 24.183295] __rtnl_newlink+0x63d/0x9f0 [ 24.183301] ? kmem_cache_alloc_node_noprof+0x12b/0x360 [ 24.183308] ? kmalloc_trace_noprof+0x11e/0x350 [ 24.183312] ? rtnl_newlink+0x2e/0x70 [ 24.183316] rtnl_newlink+0x47/0x70 [ 24.183320] rtnetlink_rcv_msg+0x152/0x400 [ 24.183324] ? __netlink_sendskb+0x68/0x90 [ 24.183329] ? netlink_unicast+0x237/0x290 [ 24.183333] ? __pfx_rtnetlink_rcv_msg+0x10/0x10 [ 24.183336] netlink_rcv_skb+0x5b/0x110 [ 24.183343] netlink_unicast+0x1a4/0x290 [ 24.183347] netlink_sendmsg+0x222/0x4a0 [ 24.183350] ? proc_get_long.constprop.0+0x116/0x210 [ 24.183358] ____sys_sendmsg+0x379/0x3b0 [ 24.183363] ? copy_msghdr_from_user+0x6d/0xb0 [ 24.183368] ___sys_sendmsg+0x86/0xe0 [ 24.183372] ? addrconf_sysctl_forward+0xf3/0x270 [ 24.183378] ? _copy_from_iter+0x8b/0x570 [ 24.183384] ? __pfx_addrconf_sysctl_forward+0x10/0x10 [ 24.183388] ? _raw_spin_unlock+0x19/0x50 [ 24.183392] ? proc_sys_call_handler+0xf3/0x2f0 [ 24.183397] ? trace_hardirqs_on+0x29/0x90 [ 24.183401] ? __fdget+0xc2/0xf0 [ 24.183405] __sys_sendmsg+0x5b/0xc0 [ 24.183410] ? syscall_trace_enter+0x110/0x1b0 [ 24.183416] do_syscall_64+0x64/0x150 [ 24.183423] entry_SYSCALL_64_after_hwframe+0x76/0x7e I have bisected the error to this commit. Reverting it caused no new or perceivable issues on both the MacBook and a Zen4-based laptop. Revert this commit as a workaround. This reverts commit aa162aa4aa383a0a714b1c36e8fcc77612ddd1a2. Upstream report: https://bugzilla.kernel.org/show_bug.cgi?id=219390 Signed-off-by: Mingcong Bai Bug: https://lore.kernel.org/all/b8da4aec-4cca-4eb0-ba87-5f8641aa2ca9@leemhuis.info/ Signed-off-by: Kexy Biscuit --- kernel/rcu/tree.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index ff98233d4aa59f..b71dcf77d16616 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -5227,15 +5227,11 @@ void rcutree_migrate_callbacks(int cpu) struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu); bool needwake; - if (rcu_rdp_is_offloaded(rdp)) - return; - - raw_spin_lock_irqsave(&rcu_state.barrier_lock, flags); - if (rcu_segcblist_empty(&rdp->cblist)) { - raw_spin_unlock_irqrestore(&rcu_state.barrier_lock, flags); + if (rcu_rdp_is_offloaded(rdp) || + rcu_segcblist_empty(&rdp->cblist)) return; /* No callbacks to migrate. */ - } + raw_spin_lock_irqsave(&rcu_state.barrier_lock, flags); WARN_ON_ONCE(rcu_rdp_cpu_online(rdp)); rcu_barrier_entrain(rdp); my_rdp = this_cpu_ptr(&rcu_data);