On Fri, Jul 03, 2026 at 04:48:10PM +0800, Jun Miao wrote:
> When the kernel boots from kexec, the EPC pages may have a stale state.
> The kernel sanitizes all EPC pages to reset them to a clean state before
> their first use in any enclave. The EPC size could be several GBs and
> resetting them could take a significant amount of time. Because of that,
> the kernel performs the reset in a loop through a kernel thread ksgxd() at
> early boot, and there's a cond_resched() after resetting each EPC page.
>
> This is fine in most cases, but becomes a problem when there's other kernel
> code waiting for an RCU-Tasks grace period but the cond_resched() in
> ksgxd() never triggers rescheduling. Because cond_resched() doesn't report
> a quiescent state when it doesn't trigger rescheduling, the thread that is
> waiting for an RCU-Tasks grace period will wait until all EPC pages are
> reset.
You should probably also say why cond_resched() does not trigger
rescheduling. E.g., put that and delete the first sentence as
it tells nothing and is confusing.
>
> For instance, BPF LSM subsystem can invoke synchronize_rcu_tasks() at
> kernel boot time. A VM with a large EPC assigned and BPF LSM enabled can
> take a long time to boot, with a call trace triggered:
>
> rcu_tasks_wait_gp: rcu_tasks grace period number 1 (since boot) is
> 130631 jiffies old.
> INFO: task systemd:1 blocked for more than 122 seconds.
> ...
> task:systemd state:D stack:0 pid:1 tpid:1 ppid:0 flags:0x00000002
> Call Trace:
> ...
> schedule_timeout+0x157/0x170
> wait_for_completion+0x88/0x150
> __wait_rcu_gp+0x17e/0x190
> synchronize_rcu_tasks_generic+0x64/0x60
> ...
> synchronize_rcu_tasks+0x15/0x20
> register_ftrace_direct+0x31f/0x350
> ...
> bpf_trampoline_link_prog+0x33/0x60
> bpf_tracing_prog_attach+0x3c5/0x5f0
>
> Replace cond_resched() with cond_resched_tasks_rcu_qs() which explicitly
> reports quiescent state regardless of whether actual rescheduling is
> triggered. Resetting all EPC pages in ksgxd() isn't performance critical
> so the extra cost of cond_resched_tasks_rcu_qs() isn't a problem.
>
> Tests showed this reduced the VM kernel boot time from ~50s to ~700ms.
>
> Fixes: e7e0545299d8 ("x86/sgx: Initialize metadata for Enclave Page Cache
> (EPC) sections")
> Suggested-by: Kai Huang <[email protected]>
> Co-developed-by: Fan Du <[email protected]>
> Signed-off-by: Fan Du <[email protected]>
> Signed-off-by: Jun Miao <[email protected]>
> Tested-by: Challvy Tee <[email protected]>
> Reviewed-by: Kai Huang <[email protected]>
> Link: https://github.com/systemd/systemd/issues/40423
>
> ---
> v1 -> v2:
> - Clarify the RCU Tasks stall root cause.
> - Use cond_resched_rcu_qs() following the Kai`s suggestion.
>
> v2 -> v3:
> - cee439398933 ("rcu: Rename cond_resched_rcu_qs() to
> cond_resched_tasks_rcu_qs()")
>
> v3 -> v4:
> - Trim down/rewrite changelog following Kai`s suggestion.
>
> v4 -> v5:
> - Change the title, not state the problem directly
> - Corrected spelling and grammatical errors by Kai
> - Add "Reviewed-by: Kai Huang"
>
> ---
> arch/x86/kernel/cpu/sgx/main.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c
> index 4505f808af5e..7d2f57663177 100644
> --- a/arch/x86/kernel/cpu/sgx/main.c
> +++ b/arch/x86/kernel/cpu/sgx/main.c
> @@ -106,7 +106,7 @@ static unsigned long __sgx_sanitize_pages(struct
> list_head *dirty_page_list)
> left_dirty++;
> }
>
/*
* Pretend to be in the quiescent state given the possiblity of
* an ongoing grace period:
*/
> - cond_resched();
> + cond_resched_tasks_rcu_qs();
> }
>
> list_splice(&dirty, dirty_page_list);
> --
> 2.32.0
>
BR, Jarkko
