Re: [PATCH 1/3] entry: move exit to usermode functions to header file

From: Sven Schnelle
Date: Mon Dec 18 2023 - 02:47:03 EST


Hi Thomas,

Thomas Gleixner <tglx@xxxxxxxxxxxxx> writes:

> On Tue, Dec 05 2023 at 14:30, Sven Schnelle wrote:
>> +/**
>> + * exit_to_user_mode_loop - do any pending work before leaving to user space
>> + */
>> +static __always_inline unsigned long exit_to_user_mode_loop(struct pt_regs *regs,
>> + unsigned long ti_work)
>> +{
>> + /*
>> + * Before returning to user space ensure that all pending work
>> + * items have been completed.
>> + */
>> + while (ti_work & EXIT_TO_USER_MODE_WORK) {
>> +
>> + local_irq_enable_exit_to_user(ti_work);
>> +
>> + if (ti_work & _TIF_NEED_RESCHED)
>> + schedule();
>> +
>> + if (ti_work & _TIF_UPROBE)
>> + uprobe_notify_resume(regs);
>> +
>> + if (ti_work & _TIF_PATCH_PENDING)
>> + klp_update_patch_state(current);
>> +
>> + if (ti_work & (_TIF_SIGPENDING | _TIF_NOTIFY_SIGNAL))
>> + arch_do_signal_or_restart(regs);
>> +
>> + if (ti_work & _TIF_NOTIFY_RESUME)
>> + resume_user_mode_work(regs);
>> +
>> + /* Architecture specific TIF work */
>> + arch_exit_to_user_mode_work(regs, ti_work);
>> +
>> + /*
>> + * Disable interrupts and reevaluate the work flags as they
>> + * might have changed while interrupts and preemption was
>> + * enabled above.
>> + */
>> + local_irq_disable_exit_to_user();
>> +
>> + /* Check if any of the above work has queued a deferred wakeup */
>> + tick_nohz_user_enter_prepare();
>> +
>> + ti_work = read_thread_flags();
>> + }
>> +
>> + /* Return the latest work state for arch_exit_to_user_mode() */
>> + return ti_work;
>> +}
>
> I'm not really sure about this part. exit_to_user_mode_loop() is the
> slowpath when a TIF work flag is set. I can see the benefit on the
> fastpath functions which are way smaller.

Indeed, the main performance improvement comes from inlining the small
functions. As Peter mentioned the same, i sent out a v2 which doesn't
move exit_to_user_mode_loop().

Thanks!
Sven