//.........这里部分代码省略.........
        mark_buf[mark_len] = saved_char;
    }

    /*
     * Copy the output into log_buf.  If the caller didn't provide
     * appropriate log level tags, we insert them here
     */
    for (p = printk_buf; *p; p++) {
        if (log_level_unknown) {
            /* log_level_unknown signals the start of a new line */
            if (printk_time) {
                int loglev_char;
                char tbuf[50], *tp;
                unsigned tlen;
                unsigned long long t;
                unsigned long nanosec_rem;

                /*
                 * force the log level token to be
                 * before the time output.
                 */
                if (p[0] == '<' && p[1] >='0' &&
                        p[1] <= '7' && p[2] == '>') {
                    loglev_char = p[1];
                    p += 3;
                    printed_len -= 3;
                } else {
                    loglev_char = default_message_loglevel
                                  + '0';
                }
                t = printk_clock();
                nanosec_rem = do_div(t, 1000000000);
                tlen = sprintf(tbuf,
                               "<%c>[%5lu.%06lu] ",
                               loglev_char,
                               (unsigned long)t,
                               nanosec_rem/1000);

                for (tp = tbuf; tp < tbuf + tlen; tp++)
                    emit_log_char(*tp);
                printed_len += tlen;
            } else {
                if (p[0] != '<' || p[1] < '0' ||
                        p[1] > '7' || p[2] != '>') {
                    emit_log_char('<');
                    emit_log_char(default_message_loglevel
                                  + '0');
                    emit_log_char('>');
                    printed_len += 3;
                }
            }
            log_level_unknown = 0;
            if (!*p)
                break;
        }
        emit_log_char(*p);
        if (*p == '\n')
            log_level_unknown = 1;
    }

    if (!down_trylock(&console_sem)) {
        /*
         * We own the drivers.  We can drop the spinlock and
         * let release_console_sem() print the text, maybe ...
         */
        console_locked = 1;
        printk_cpu = UINT_MAX;
        spin_unlock(&logbuf_lock);

        /*
         * Console drivers may assume that per-cpu resources have
         * been allocated. So unless they're explicitly marked as
         * being able to cope (CON_ANYTIME) don't call them until
         * this CPU is officially up.
         */
        if (cpu_online(smp_processor_id()) || have_callable_console()) {
            console_may_schedule = 0;
            release_console_sem();
        } else {
            /* Release by hand to avoid flushing the buffer. */
            console_locked = 0;
            up(&console_sem);
        }
        lockdep_on();
        raw_local_irq_restore(flags);
    } else {
        /*
         * Someone else owns the drivers.  We drop the spinlock, which
         * allows the semaphore holder to proceed and to call the
         * console drivers with the output which we just produced.
         */
        printk_cpu = UINT_MAX;
        spin_unlock(&logbuf_lock);
        lockdep_on();
        raw_local_irq_restore(flags);
    }

    preempt_enable();
    return printed_len;
}

//.........这里部分代码省略.........
	   monitor process attaching to us before the non-preemptible
	   perfctr update step. Therefore we cannot store the copy
	   in the perfctr object itself. */
	control = kmalloc(sizeof(*control), GFP_USER);
	if (!control) {
		return -ENOMEM;
	}

	err = -EINVAL;
	if (argbytes > sizeof *control) {
		goto out_kfree;
	}

	err = -EFAULT;
	if (copy_from_user(control, argp, argbytes)) {
		goto out_kfree;
	}

	if (argbytes < sizeof *control)
		memset((char*)control + argbytes, 0, sizeof *control - argbytes);

	// figure out what is happening in the following 'if' loop

	if (control->cpu_control.nractrs || control->cpu_control.nrictrs) {
		cpumask_t old_mask, new_mask;

		old_mask = tsk->cpus_allowed;
		cpus_andnot(new_mask, old_mask, perfctr_cpus_forbidden_mask);

		err = -EINVAL;
		if (cpus_empty(new_mask)) {
			goto out_kfree;
		}
		if (!cpus_equal(new_mask, old_mask))
			set_cpus_allowed(tsk, new_mask);
	}

	/* PREEMPT note: preemption is disabled over the entire
	   region since we're updating an active perfctr. */
	preempt_disable();

	// the task whose control register I am changing might actually be
	// in suspended state. That can happen when the other is executing
	// under the control of another task as in the case of debugging
	// or ptrace. However, if the write_control is done for the current
	// executing process, first suspend them and then do the update
	// why are we resetting 'perfctr->cpu_state.cstatus' ?

	if (IS_RUNNING(perfctr)) {
		if (tsk == current)
			vperfctr_suspend(perfctr);
	
		// not sure why we are zeroing out the following explicitly
		perfctr->cpu_state.cstatus = 0;
		vperfctr_clear_iresume_cstatus(perfctr);
	}

	// coying the user-specified control values to 'state'
	perfctr->cpu_state.control = control->cpu_control;

	/* remote access note: perfctr_cpu_update_control() is ok */
	err = perfctr_cpu_update_control(&perfctr->cpu_state, 0);
	if (err < 0) {
		goto out;
	}
	next_cstatus = perfctr->cpu_state.cstatus;
	if (!perfctr_cstatus_enabled(next_cstatus))
		goto out;

	/* XXX: validate si_signo? */
	perfctr->si_signo = control->si_signo;

	if (!perfctr_cstatus_has_tsc(next_cstatus))
		perfctr->cpu_state.tsc_sum = 0;

	nrctrs = perfctr_cstatus_nrctrs(next_cstatus);
	for(i = 0; i < nrctrs; ++i)
		if (!(control->preserve & (1<<i)))
			perfctr->cpu_state.pmc[i].sum = 0;

	// I am not sure why we are removing the inheritance just because
	// we updated the control information. True, because the children might
	// be performing something else. So, the control will have to be set
	// before spawning any children

	spin_lock(&perfctr->children_lock);
	perfctr->inheritance_id = new_inheritance_id();
	memset(&perfctr->children, 0, sizeof perfctr->children);
	spin_unlock(&perfctr->children_lock);

	if (tsk == current) {
		vperfctr_resume(perfctr);
	}

 out:
	preempt_enable();
 out_kfree:
	kfree(control);
	return err;
}

/*
 * Package up a bounce condition.
 */
void VFP_bounce(u32 trigger, u32 fpexc, struct pt_regs *regs)
{
	u32 fpscr, orig_fpscr, fpsid, exceptions;

	pr_debug("VFP: bounce: trigger %08x fpexc %08x\n", trigger, fpexc);

	/*
	 * At this point, FPEXC can have the following configuration:
	 *
	 *  EX DEX IXE
	 *  0   1   x   - synchronous exception
	 *  1   x   0   - asynchronous exception
	 *  1   x   1   - sychronous on VFP subarch 1 and asynchronous on later
	 *  0   0   1   - synchronous on VFP9 (non-standard subarch 1
	 *                implementation), undefined otherwise
	 *
	 * Clear various bits and enable access to the VFP so we can
	 * handle the bounce.
	 */
	fmxr(FPEXC, fpexc & ~(FPEXC_EX|FPEXC_DEX|FPEXC_FP2V|FPEXC_VV|FPEXC_TRAP_MASK));

	fpsid = fmrx(FPSID);
	orig_fpscr = fpscr = fmrx(FPSCR);

	/*
	 * Check for the special VFP subarch 1 and FPSCR.IXE bit case
	 */
	if ((fpsid & FPSID_ARCH_MASK) == (1 << FPSID_ARCH_BIT)
	    && (fpscr & FPSCR_IXE)) {
		/*
		 * Synchronous exception, emulate the trigger instruction
		 */
		goto emulate;
	}

	if (fpexc & FPEXC_EX) {
#ifndef CONFIG_CPU_FEROCEON
		/*
		 * Asynchronous exception. The instruction is read from FPINST
		 * and the interrupted instruction has to be restarted.
		 */
		trigger = fmrx(FPINST);
		regs->ARM_pc -= 4;
#endif
	} else if (!(fpexc & FPEXC_DEX)) {
		/*
		 * Illegal combination of bits. It can be caused by an
		 * unallocated VFP instruction but with FPSCR.IXE set and not
		 * on VFP subarch 1.
		 */
		 vfp_raise_exceptions(VFP_EXCEPTION_ERROR, trigger, fpscr, regs);
		goto exit;
	}

	/*
	 * Modify fpscr to indicate the number of iterations remaining.
	 * If FPEXC.EX is 0, FPEXC.DEX is 1 and the FPEXC.VV bit indicates
	 * whether FPEXC.VECITR or FPSCR.LEN is used.
	 */
	if (fpexc & (FPEXC_EX | FPEXC_VV)) {
		u32 len;

		len = fpexc + (1 << FPEXC_LENGTH_BIT);

		fpscr &= ~FPSCR_LENGTH_MASK;
		fpscr |= (len & FPEXC_LENGTH_MASK) << (FPSCR_LENGTH_BIT - FPEXC_LENGTH_BIT);
	}

	/*
	 * Handle the first FP instruction.  We used to take note of the
	 * FPEXC bounce reason, but this appears to be unreliable.
	 * Emulate the bounced instruction instead.
	 */
	exceptions = vfp_emulate_instruction(trigger, fpscr, regs);
	if (exceptions)
		vfp_raise_exceptions(exceptions, trigger, orig_fpscr, regs);

	/*
	 * If there isn't a second FP instruction, exit now. Note that
	 * the FPEXC.FP2V bit is valid only if FPEXC.EX is 1.
	 */
	if ((fpexc & (FPEXC_EX | FPEXC_FP2V)) != (FPEXC_EX | FPEXC_FP2V))
		goto exit;

	/*
	 * The barrier() here prevents fpinst2 being read
	 * before the condition above.
	 */
	barrier();
	trigger = fmrx(FPINST2);

 emulate:
	exceptions = vfp_emulate_instruction(trigger, orig_fpscr, regs);
	if (exceptions)
		vfp_raise_exceptions(exceptions, trigger, orig_fpscr, regs);
 exit:
	preempt_enable();
//.........这里部分代码省略.........

int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
{
	struct address_space *swap_space;
	struct file *swap_storage;
	struct page *from_page;
	struct page *to_page;
	void *from_virtual;
	void *to_virtual;
	int i;
	int ret = -ENOMEM;

	BUG_ON(ttm->state != tt_unbound && ttm->state != tt_unpopulated);
	BUG_ON(ttm->caching_state != tt_cached);

	/*
	 * For user buffers, just unpin the pages, as there should be
	 * vma references.
	 */

	if (ttm->page_flags & TTM_PAGE_FLAG_USER) {
		ttm_tt_free_user_pages(ttm);
		ttm->page_flags |= TTM_PAGE_FLAG_SWAPPED;
		ttm->swap_storage = NULL;
		return 0;
	}

	if (!persistant_swap_storage) {
		swap_storage = shmem_file_setup("ttm swap",
						ttm->num_pages << PAGE_SHIFT,
						0);
		if (unlikely(IS_ERR(swap_storage))) {
			printk(KERN_ERR "Failed allocating swap storage.\n");
			return PTR_ERR(swap_storage);
		}
	} else
		swap_storage = persistant_swap_storage;

	swap_space = swap_storage->f_path.dentry->d_inode->i_mapping;

	for (i = 0; i < ttm->num_pages; ++i) {
		from_page = ttm->pages[i];
		if (unlikely(from_page == NULL))
			continue;
		to_page = read_mapping_page(swap_space, i, NULL);
		if (unlikely(IS_ERR(to_page))) {
			ret = PTR_ERR(to_page);
			goto out_err;
		}
		preempt_disable();
		from_virtual = kmap_atomic(from_page, KM_USER0);
		to_virtual = kmap_atomic(to_page, KM_USER1);
		memcpy(to_virtual, from_virtual, PAGE_SIZE);
		kunmap_atomic(to_virtual, KM_USER1);
		kunmap_atomic(from_virtual, KM_USER0);
		preempt_enable();
		set_page_dirty(to_page);
		mark_page_accessed(to_page);
		page_cache_release(to_page);
	}

	ttm_tt_free_alloced_pages(ttm);
	ttm->swap_storage = swap_storage;
	ttm->page_flags |= TTM_PAGE_FLAG_SWAPPED;
	if (persistant_swap_storage)
		ttm->page_flags |= TTM_PAGE_FLAG_PERSISTANT_SWAP;

	return 0;
out_err:
	if (!persistant_swap_storage)
		fput(swap_storage);

	return ret;
}

int
trace_selftest_startup_preemptirqsoff(struct tracer *trace, struct trace_array *tr)
{
	unsigned long save_max = tracing_max_latency;
	unsigned long count;
	int ret;

	/*
	 * Now that the big kernel lock is no longer preemptable,
	 * and this is called with the BKL held, it will always
	 * fail. If preemption is already disabled, simply
	 * pass the test. When the BKL is removed, or becomes
	 * preemptible again, we will once again test this,
	 * so keep it in.
	 */
	if (preempt_count()) {
		printk(KERN_CONT "can not test ... force ");
		return 0;
	}

	/* start the tracing */
	ret = trace->init(tr);
	if (ret) {
		warn_failed_init_tracer(trace, ret);
		goto out;
	}

	/* reset the max latency */
	tracing_max_latency = 0;

	/* disable preemption and interrupts for a bit */
	preempt_disable();
	local_irq_disable();
	udelay(100);
	preempt_enable();
	/* reverse the order of preempt vs irqs */
	local_irq_enable();

	/* stop the tracing. */
	tracing_stop();
	/* check both trace buffers */
	ret = trace_test_buffer(tr, NULL);
	if (ret) {
		tracing_start();
		goto out;
	}

	ret = trace_test_buffer(&max_tr, &count);
	if (ret) {
		tracing_start();
		goto out;
	}

	if (!ret && !count) {
		printk(KERN_CONT ".. no entries found ..");
		ret = -1;
		tracing_start();
		goto out;
	}

	/* do the test by disabling interrupts first this time */
	tracing_max_latency = 0;
	tracing_start();
	preempt_disable();
	local_irq_disable();
	udelay(100);
	preempt_enable();
	/* reverse the order of preempt vs irqs */
	local_irq_enable();

	/* stop the tracing. */
	tracing_stop();
	/* check both trace buffers */
	ret = trace_test_buffer(tr, NULL);
	if (ret)
		goto out;

	ret = trace_test_buffer(&max_tr, &count);

	if (!ret && !count) {
		printk(KERN_CONT ".. no entries found ..");
		ret = -1;
		goto out;
	}

 out:
	trace->reset(tr);
	tracing_start();
	tracing_max_latency = save_max;

	return ret;
}

static void gru_free_cpu_resources(void *cb, void *dsr)
{
	preempt_enable();
}

static int mcsdl_download(const UINT8 *pBianry, const UINT16 unLength,
		INT8 IdxNum)
{
	int nRet;

	//---------------------------------
	// Check Binary Size
	//---------------------------------
	if (unLength >= MELFAS_FIRMWARE_MAX_SIZE)
	{

		nRet = MCSDL_RET_PROGRAM_SIZE_IS_WRONG;
		goto MCSDL_DOWNLOAD_FINISH;
	}

#if MELFAS_ENABLE_DBG_PROGRESS_PRINT
	printk(" - Starting download...\n");
#endif

	//---------------------------------
	// Make it ready
	//---------------------------------
#if MELFAS_ENABLE_DBG_PROGRESS_PRINT
	printk(" > Ready\n");
#endif

	mcsdl_set_ready();

	//---------------------------------
	// Erase Flash
	//---------------------------------
#if MELFAS_ENABLE_DBG_PROGRESS_PRINT
	printk(" > Erase\n");
#endif

	printk(" > Erase - Start \n");
	preempt_disable();
	nRet = mcsdl_erase_flash(IdxNum);
	preempt_enable();
	printk(" > Erase- End \n");

	if (nRet != MCSDL_RET_SUCCESS)
		goto MCSDL_DOWNLOAD_FINISH;

	//goto MCSDL_DOWNLOAD_FINISH; //lcs_test

	//---------------------------------
	// Program Flash
	//---------------------------------
#if MELFAS_ENABLE_DBG_PROGRESS_PRINT
	printk(" > Program   ");
#endif

	preempt_disable();
	nRet = mcsdl_program_flash((UINT8*) pBianry, (UINT16) unLength, IdxNum);
	preempt_enable();
	if (nRet != MCSDL_RET_SUCCESS)
		goto MCSDL_DOWNLOAD_FINISH;
	//---------------------------------
	// Verify flash
	//---------------------------------

#if MELFAS_ENABLE_DBG_PROGRESS_PRINT
	printk(" > Verify    ");
#endif

	preempt_disable();
	nRet = mcsdl_verify_flash((UINT8*) pBianry, (UINT16) unLength, IdxNum);
	preempt_enable();

	if (nRet != MCSDL_RET_SUCCESS)
		goto MCSDL_DOWNLOAD_FINISH;

	nRet = MCSDL_RET_SUCCESS;

	MCSDL_DOWNLOAD_FINISH:

#if MELFAS_ENABLE_DBG_PRINT
	mcsdl_print_result( nRet ); // Show result
#endif

#if MELFAS_ENABLE_DBG_PROGRESS_PRINT
	printk(" > Rebooting\n");
	printk(" - Fin.\n\n");
#endif

	mcsdl_reboot_mcs();

	return nRet;
}

static inline void __unlock_kernel(void)
{
	_raw_spin_unlock(&kernel_flag);
	preempt_enable();
}

/**
 * hrtimer_start_range_ns - (re)start an hrtimer on the current CPU
 * @timer:	the timer to be added
 * @tim:	expiry time
 * @delta_ns:	"slack" range for the timer
 * @mode:	expiry mode: absolute (HRTIMER_ABS) or relative (HRTIMER_REL)
 *
 * Returns:
 *  0 on success
 *  1 when the timer was active
 */
int
hrtimer_start_range_ns(struct hrtimer *timer, ktime_t tim, unsigned long delta_ns,
			const enum hrtimer_mode mode)
{
	struct hrtimer_clock_base *base, *new_base;
	unsigned long flags;
	int ret, raise;

	base = lock_hrtimer_base(timer, &flags);

	/* Remove an active timer from the queue: */
	ret = remove_hrtimer(timer, base);

	/* Switch the timer base, if necessary: */
	new_base = switch_hrtimer_base(timer, base);

	if (mode == HRTIMER_MODE_REL) {
		tim = ktime_add_safe(tim, new_base->get_time());
		/*
		 * CONFIG_TIME_LOW_RES is a temporary way for architectures
		 * to signal that they simply return xtime in
		 * do_gettimeoffset(). In this case we want to round up by
		 * resolution when starting a relative timer, to avoid short
		 * timeouts. This will go away with the GTOD framework.
		 */
#ifdef CONFIG_TIME_LOW_RES
		tim = ktime_add_safe(tim, base->resolution);
#endif
	}

	hrtimer_set_expires_range_ns(timer, tim, delta_ns);

	timer_stats_hrtimer_set_start_info(timer);

	/*
	 * Only allow reprogramming if the new base is on this CPU.
	 * (it might still be on another CPU if the timer was pending)
	 */
	enqueue_hrtimer(timer, new_base,
			new_base->cpu_base == &__get_cpu_var(hrtimer_bases));

	/*
	 * The timer may be expired and moved to the cb_pending
	 * list. We can not raise the softirq with base lock held due
	 * to a possible deadlock with runqueue lock.
	 */
	raise = timer->state == HRTIMER_STATE_PENDING;

	/*
	 * We use preempt_disable to prevent this task from migrating after
	 * setting up the softirq and raising it. Otherwise, if me migrate
	 * we will raise the softirq on the wrong CPU.
	 */
	preempt_disable();

	unlock_hrtimer_base(timer, &flags);

	if (raise)
		hrtimer_raise_softirq();
	preempt_enable();

	return ret;
}

static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
				enum kfd_preempt_type reset_type,
				unsigned int utimeout, uint32_t pipe_id,
				uint32_t queue_id)
{
	struct amdgpu_device *adev = get_amdgpu_device(kgd);
	uint32_t temp;
	enum hqd_dequeue_request_type type;
	unsigned long flags, end_jiffies;
	int retry;
	struct vi_mqd *m = get_mqd(mqd);

	if (adev->in_gpu_reset)
		return -EIO;

	acquire_queue(kgd, pipe_id, queue_id);

	if (m->cp_hqd_vmid == 0)
		WREG32_FIELD(RLC_CP_SCHEDULERS, scheduler1, 0);

	switch (reset_type) {
	case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN:
		type = DRAIN_PIPE;
		break;
	case KFD_PREEMPT_TYPE_WAVEFRONT_RESET:
		type = RESET_WAVES;
		break;
	default:
		type = DRAIN_PIPE;
		break;
	}

	/* Workaround: If IQ timer is active and the wait time is close to or
	 * equal to 0, dequeueing is not safe. Wait until either the wait time
	 * is larger or timer is cleared. Also, ensure that IQ_REQ_PEND is
	 * cleared before continuing. Also, ensure wait times are set to at
	 * least 0x3.
	 */
	local_irq_save(flags);
	preempt_disable();
	retry = 5000; /* wait for 500 usecs at maximum */
	while (true) {
		temp = RREG32(mmCP_HQD_IQ_TIMER);
		if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, PROCESSING_IQ)) {
			pr_debug("HW is processing IQ\n");
			goto loop;
		}
		if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, ACTIVE)) {
			if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, RETRY_TYPE)
					== 3) /* SEM-rearm is safe */
				break;
			/* Wait time 3 is safe for CP, but our MMIO read/write
			 * time is close to 1 microsecond, so check for 10 to
			 * leave more buffer room
			 */
			if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, WAIT_TIME)
					>= 10)
				break;
			pr_debug("IQ timer is active\n");
		} else
			break;
loop:
		if (!retry) {
			pr_err("CP HQD IQ timer status time out\n");
			break;
		}
		ndelay(100);
		--retry;
	}
	retry = 1000;
	while (true) {
		temp = RREG32(mmCP_HQD_DEQUEUE_REQUEST);
		if (!(temp & CP_HQD_DEQUEUE_REQUEST__IQ_REQ_PEND_MASK))
			break;
		pr_debug("Dequeue request is pending\n");

		if (!retry) {
			pr_err("CP HQD dequeue request time out\n");
			break;
		}
		ndelay(100);
		--retry;
	}
	local_irq_restore(flags);
	preempt_enable();

	WREG32(mmCP_HQD_DEQUEUE_REQUEST, type);

	end_jiffies = (utimeout * HZ / 1000) + jiffies;
	while (true) {
		temp = RREG32(mmCP_HQD_ACTIVE);
		if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK))
			break;
		if (time_after(jiffies, end_jiffies)) {
			pr_err("cp queue preemption time out.\n");
			release_queue(kgd);
			return -ETIME;
		}
		usleep_range(500, 1000);
	}
//.........这里部分代码省略.........

/*
 * Fill the partition reserved page with the information needed by
 * other partitions to discover we are alive and establish initial
 * communications.
 */
struct xpc_rsvd_page *
xpc_rsvd_page_init(void)
{
    struct xpc_rsvd_page *rp;
    AMO_t *amos_page;
    u64 rp_pa, nasid_array = 0;
    int i, ret;

    /* get the local reserved page's address */

    preempt_disable();
    rp_pa = xpc_get_rsvd_page_pa(cpuid_to_nasid(smp_processor_id()));
    preempt_enable();
    if (rp_pa == 0) {
        dev_err(xpc_part, "SAL failed to locate the reserved page\n");
        return NULL;
    }
    rp = (struct xpc_rsvd_page *)__va(rp_pa);

    if (rp->partid != sn_partition_id) {
        dev_err(xpc_part, "the reserved page's partid of %d should be "
            "%d\n", rp->partid, sn_partition_id);
        return NULL;
    }

    rp->version = XPC_RP_VERSION;

    /* establish the actual sizes of the nasid masks */
    if (rp->SAL_version == 1) {
        /* SAL_version 1 didn't set the nasids_size field */
        rp->nasids_size = 128;
    }
    xp_nasid_mask_bytes = rp->nasids_size;
    xp_nasid_mask_words = xp_nasid_mask_bytes / 8;

    /* setup the pointers to the various items in the reserved page */
    xpc_part_nasids = XPC_RP_PART_NASIDS(rp);
    xpc_mach_nasids = XPC_RP_MACH_NASIDS(rp);
    xpc_vars = XPC_RP_VARS(rp);
    xpc_vars_part = XPC_RP_VARS_PART(rp);

    /*
     * Before clearing xpc_vars, see if a page of AMOs had been previously
     * allocated. If not we'll need to allocate one and set permissions
     * so that cross-partition AMOs are allowed.
     *
     * The allocated AMO page needs MCA reporting to remain disabled after
     * XPC has unloaded.  To make this work, we keep a copy of the pointer
     * to this page (i.e., amos_page) in the struct xpc_vars structure,
     * which is pointed to by the reserved page, and re-use that saved copy
     * on subsequent loads of XPC. This AMO page is never freed, and its
     * memory protections are never restricted.
     */
    amos_page = xpc_vars->amos_page;
    if (amos_page == NULL) {
        amos_page = (AMO_t *)TO_AMO(uncached_alloc_page(0, 1));
        if (amos_page == NULL) {
            dev_err(xpc_part, "can't allocate page of AMOs\n");
            return NULL;
        }

        /*
         * Open up AMO-R/W to cpu.  This is done for Shub 1.1 systems
         * when xpc_allow_IPI_ops() is called via xpc_hb_init().
         */
        if (!enable_shub_wars_1_1()) {
            ret = sn_change_memprotect(ia64_tpa((u64)amos_page),
                           PAGE_SIZE,
                           SN_MEMPROT_ACCESS_CLASS_1,
                           &nasid_array);
            if (ret != 0) {
                dev_err(xpc_part, "can't change memory "
                    "protections\n");
                uncached_free_page(__IA64_UNCACHED_OFFSET |
                           TO_PHYS((u64)amos_page), 1);
                return NULL;
            }
        }
    } else if (!IS_AMO_ADDRESS((u64)amos_page)) {
        /*
         * EFI's XPBOOT can also set amos_page in the reserved page,
         * but it happens to leave it as an uncached physical address
         * and we need it to be an uncached virtual, so we'll have to
         * convert it.
         */
        if (!IS_AMO_PHYS_ADDRESS((u64)amos_page)) {
            dev_err(xpc_part, "previously used amos_page address "
                "is bad = 0x%p\n", (void *)amos_page);
            return NULL;
        }
        amos_page = (AMO_t *)TO_AMO((u64)amos_page);
    }

    /* clear xpc_vars */
    memset(xpc_vars, 0, sizeof(struct xpc_vars));
//.........这里部分代码省略.........

static int clamp_thread(void *arg)
{
	int cpunr = (unsigned long)arg;
	DEFINE_TIMER(wakeup_timer, noop_timer, 0, 0);
	static const struct sched_param param = {
		.sched_priority = MAX_USER_RT_PRIO/2,
	};
	unsigned int count = 0;
	unsigned int target_ratio;

	set_bit(cpunr, cpu_clamping_mask);
	set_freezable();
	init_timer_on_stack(&wakeup_timer);
	sched_setscheduler(current, SCHED_FIFO, &param);

	while (true == clamping && !kthread_should_stop() &&
		cpu_online(cpunr)) {
		int sleeptime;
		unsigned long target_jiffies;
		unsigned int guard;
		unsigned int compensation = 0;
		int interval; /* jiffies to sleep for each attempt */
		unsigned int duration_jiffies = msecs_to_jiffies(duration);
		unsigned int window_size_now;

		try_to_freeze();
		/*
		 * make sure user selected ratio does not take effect until
		 * the next round. adjust target_ratio if user has changed
		 * target such that we can converge quickly.
		 */
		target_ratio = set_target_ratio;
		guard = 1 + target_ratio/20;
		window_size_now = window_size;
		count++;

		/*
		 * systems may have different ability to enter package level
		 * c-states, thus we need to compensate the injected idle ratio
		 * to achieve the actual target reported by the HW.
		 */
		compensation = get_compensation(target_ratio);
		interval = duration_jiffies*100/(target_ratio+compensation);

		/* align idle time */
		target_jiffies = roundup(jiffies, interval);
		sleeptime = target_jiffies - jiffies;
		if (sleeptime <= 0)
			sleeptime = 1;
		schedule_timeout_interruptible(sleeptime);
		/*
		 * only elected controlling cpu can collect stats and update
		 * control parameters.
		 */
		if (cpunr == control_cpu && !(count%window_size_now)) {
			should_skip =
				powerclamp_adjust_controls(target_ratio,
							guard, window_size_now);
			smp_mb();
		}

		if (should_skip)
			continue;

		target_jiffies = jiffies + duration_jiffies;
		mod_timer(&wakeup_timer, target_jiffies);
		if (unlikely(local_softirq_pending()))
			continue;
		/*
		 * stop tick sched during idle time, interrupts are still
		 * allowed. thus jiffies are updated properly.
		 */
		preempt_disable();
		/* mwait until target jiffies is reached */
		while (time_before(jiffies, target_jiffies)) {
			unsigned long ecx = 1;
			unsigned long eax = target_mwait;

			/*
			 * REVISIT: may call enter_idle() to notify drivers who
			 * can save power during cpu idle. same for exit_idle()
			 */
			local_touch_nmi();
			stop_critical_timings();
			mwait_idle_with_hints(eax, ecx);
			start_critical_timings();
			atomic_inc(&idle_wakeup_counter);
		}
		preempt_enable();
	}
	del_timer_sync(&wakeup_timer);
	clear_bit(cpunr, cpu_clamping_mask);

	return 0;
}

/*
 * 1 HZ polling while clamping is active, useful for userspace
 * to monitor actual idle ratio.
 */
//.........这里部分代码省略.........

/*
 * Fill the partition reserved page with the information needed by
 * other partitions to discover we are alive and establish initial
 * communications.
 */
int
xpc_setup_rsvd_page(void)
{
	int ret;
	struct xpc_rsvd_page *rp;
	unsigned long rp_pa;
	unsigned long new_ts_jiffies;

	/* get the local reserved page's address */

	preempt_disable();
	rp_pa = xpc_get_rsvd_page_pa(xp_cpu_to_nasid(smp_processor_id()));
	preempt_enable();
	if (rp_pa == 0) {
		dev_err(xpc_part, "SAL failed to locate the reserved page\n");
		return -ESRCH;
	}
	rp = (struct xpc_rsvd_page *)__va(rp_pa);

	if (rp->SAL_version < 3) {
		/* SAL_versions < 3 had a SAL_partid defined as a u8 */
		rp->SAL_partid &= 0xff;
	}
	BUG_ON(rp->SAL_partid != xp_partition_id);

	if (rp->SAL_partid < 0 || rp->SAL_partid >= xp_max_npartitions) {
		dev_err(xpc_part, "the reserved page's partid of %d is outside "
			"supported range (< 0 || >= %d)\n", rp->SAL_partid,
			xp_max_npartitions);
		return -EINVAL;
	}

	rp->version = XPC_RP_VERSION;
	rp->max_npartitions = xp_max_npartitions;

	/* establish the actual sizes of the nasid masks */
	if (rp->SAL_version == 1) {
		/* SAL_version 1 didn't set the nasids_size field */
		rp->SAL_nasids_size = 128;
	}
	xpc_nasid_mask_nbytes = rp->SAL_nasids_size;
	xpc_nasid_mask_nlongs = BITS_TO_LONGS(rp->SAL_nasids_size *
					      BITS_PER_BYTE);

	/* setup the pointers to the various items in the reserved page */
	xpc_part_nasids = XPC_RP_PART_NASIDS(rp);
	xpc_mach_nasids = XPC_RP_MACH_NASIDS(rp);

	ret = xpc_setup_rsvd_page_sn(rp);
	if (ret != 0)
		return ret;

	/*
	 * Set timestamp of when reserved page was setup by XPC.
	 * This signifies to the remote partition that our reserved
	 * page is initialized.
	 */
	new_ts_jiffies = jiffies;
	if (new_ts_jiffies == 0 || new_ts_jiffies == rp->ts_jiffies)
		new_ts_jiffies++;
	rp->ts_jiffies = new_ts_jiffies;

	xpc_rsvd_page = rp;
	return 0;
}

//.........这里部分代码省略.........
		bdi->dirty_exceeded = 0;

	if (writeback_in_progress(bdi))
		return;		/* pdflush is already working this queue */

	/*
	 * In laptop mode, we wait until hitting the higher threshold before
	 * starting background writeout, and then write out all the way down
	 * to the lower threshold.  So slow writers cause minimal disk activity.
	 *
	 * In normal mode, we start background writeout at the lower
	 * background_thresh, to keep the amount of dirty memory low.
	 */
	if ((laptop_mode && pages_written) ||
			(!laptop_mode && (global_page_state(NR_FILE_DIRTY)
					  + global_page_state(NR_UNSTABLE_NFS)
					  > background_thresh)))
		pdflush_operation(background_writeout, 0);
}

void set_page_dirty_balance(struct page *page, int page_mkwrite)
{
	if (set_page_dirty(page) || page_mkwrite) {
		struct address_space *mapping = page_mapping(page);

		if (mapping)
			balance_dirty_pages_ratelimited(mapping);
	}
}

/**
 * balance_dirty_pages_ratelimited_nr - balance dirty memory state
 * @mapping: address_space which was dirtied
 * @nr_pages_dirtied: number of pages which the caller has just dirtied
 *
 * Processes which are dirtying memory should call in here once for each page
 * which was newly dirtied.  The function will periodically check the system's
 * dirty state and will initiate writeback if needed.
 *
 * On really big machines, get_writeback_state is expensive, so try to avoid
 * calling it too often (ratelimiting).  But once we're over the dirty memory
 * limit we decrease the ratelimiting by a lot, to prevent individual processes
 * from overshooting the limit by (ratelimit_pages) each.
 */
void balance_dirty_pages_ratelimited_nr(struct address_space *mapping,
					unsigned long nr_pages_dirtied)
{
	static DEFINE_PER_CPU(unsigned long, ratelimits) = 0;
	unsigned long ratelimit;
	unsigned long *p;

	ratelimit = ratelimit_pages;
	if (mapping->backing_dev_info->dirty_exceeded)
		ratelimit = 8;

	/*
	 * Check the rate limiting. Also, we do not want to throttle real-time
	 * tasks in balance_dirty_pages(). Period.
	 */
	preempt_disable();
	p =  &__get_cpu_var(ratelimits);
	*p += nr_pages_dirtied;
	if (unlikely(*p >= ratelimit)) {
		*p = 0;
		preempt_enable();
		balance_dirty_pages(mapping);
		return;
	}
	preempt_enable();
}
EXPORT_SYMBOL(balance_dirty_pages_ratelimited_nr);

void throttle_vm_writeout(gfp_t gfp_mask)
{
	long background_thresh;
	long dirty_thresh;

        for ( ; ; ) {
		get_dirty_limits(&background_thresh, &dirty_thresh, NULL, NULL);

                /*
                 * Boost the allowable dirty threshold a bit for page
                 * allocators so they don't get DoS'ed by heavy writers
                 */
                dirty_thresh += dirty_thresh / 10;      /* wheeee... */

                if (global_page_state(NR_UNSTABLE_NFS) +
			global_page_state(NR_WRITEBACK) <= dirty_thresh)
                        	break;
                congestion_wait(WRITE, HZ/10);

		/*
		 * The caller might hold locks which can prevent IO completion
		 * or progress in the filesystem.  So we cannot just sit here
		 * waiting for IO to complete.
		 */
		if ((gfp_mask & (__GFP_FS|__GFP_IO)) != (__GFP_FS|__GFP_IO))
			break;
        }
}

static void spe_end(void)
{
	disable_kernel_spe();
	/* reenable preemption */
	preempt_enable();
}

/*
 * All calls to this function will be optimised into tail calls. We are
 * passed a pointer to the destination which we return as required by a
 * memcpy implementation.
 */
void *exit_vmx_copy(void *dest)
{
	preempt_enable();
	return dest;
}

inline bool skb_recycler_consume(struct sk_buff *skb) {
	unsigned long flags;
	struct sk_buff_head *h;

	/* Can we recycle this skb?  If not, simply return that we cannot */
	if (unlikely(!consume_skb_can_recycle(skb, SKB_RECYCLE_MIN_SIZE,
					      SKB_RECYCLE_MAX_SIZE)))
		return false;

	/*
	 * If we can, then it will be much faster for us to recycle this one
	 * later than to allocate a new one from scratch.
	 */
	preempt_disable();
	h = &__get_cpu_var(recycle_list);
	local_irq_save(flags);
	/* Attempt to enqueue the CPU hot recycle list first */
	if (likely(skb_queue_len(h) < SKB_RECYCLE_MAX_SKBS)) {
		__skb_queue_head(h, skb);
		local_irq_restore(flags);
		preempt_enable();
		return true;
	}
#ifdef CONFIG_SKB_RECYCLER_MULTI_CPU
	h = &__get_cpu_var(recycle_spare_list);

	/* The CPU hot recycle list was full; if the spare list is also full,
	 * attempt to move the spare list to the global list for other CPUs to
	 * use.
	 */
	if (unlikely(skb_queue_len(h) >= SKB_RECYCLE_SPARE_MAX_SKBS)) {
		uint8_t cur_tail, next_tail;
		spin_lock(&glob_recycler.lock);
		cur_tail = glob_recycler.tail;
		next_tail = (cur_tail + 1) & SKB_RECYCLE_MAX_SHARED_POOLS_MASK;
		if (next_tail != glob_recycler.head) {
			struct sk_buff_head *p = &glob_recycler.pool[cur_tail];

			/* Optimized, inlined SKB List splice */
			p->next = h->next;
			h->next->prev = (struct sk_buff *)p;
			p->prev = h->prev;
			h->prev->next = (struct sk_buff *)p;
			p->qlen = SKB_RECYCLE_SPARE_MAX_SKBS;

			/* Done with global list init */
			glob_recycler.tail = next_tail;
			spin_unlock(&glob_recycler.lock);

			/* Optimized, inlined spare SKB list init */
			h->next = (struct sk_buff *)h;
			h->prev = (struct sk_buff *)h;
			h->qlen = 0;

			/* We have now cleared room in the spare; enqueue */
			__skb_queue_head(h, skb);
			local_irq_restore(flags);
			preempt_enable();
			return true;
		}
		/* We still have a full spare because the global is also full */
		spin_unlock(&glob_recycler.lock);
	} else {
		/* We have room in the spare list; enqueue to spare list */
		__skb_queue_head(h, skb);
		local_irq_restore(flags);
		preempt_enable();
		return true;
	}
#endif

	local_irq_restore(flags);
	preempt_enable();

	return false;
}

int copy_thread(unsigned long clone_flags, unsigned long usp,
	unsigned long unused, struct task_struct *p, struct pt_regs *regs)
{
	struct thread_info *ti = task_thread_info(p);
	struct pt_regs *childregs;
	unsigned long childksp;
	p->set_child_tid = p->clear_child_tid = NULL;

	childksp = (unsigned long)task_stack_page(p) + THREAD_SIZE - 32;

	preempt_disable();

	if (is_fpu_owner())
		save_fp(p);

	if (cpu_has_dsp)
		save_dsp(p);

	preempt_enable();

	/* set up new TSS. */
	childregs = (struct pt_regs *) childksp - 1;
	/*  Put the stack after the struct pt_regs.  */
	childksp = (unsigned long) childregs;
	*childregs = *regs;
	childregs->regs[7] = 0;	/* Clear error flag */

	childregs->regs[2] = 0;	/* Child gets zero as return value */

	if (childregs->cp0_status & ST0_CU0) {
		childregs->regs[28] = (unsigned long) ti;
		childregs->regs[29] = childksp;
		ti->addr_limit = KERNEL_DS;
	} else {
		childregs->regs[29] = usp;
		ti->addr_limit = USER_DS;
	}
	p->thread.reg29 = (unsigned long) childregs;
	p->thread.reg31 = (unsigned long) ret_from_fork;

	/*
	 * New tasks lose permission to use the fpu. This accelerates context
	 * switching for most programs since they don't use the fpu.
	 */
	p->thread.cp0_status = read_c0_status() & ~(ST0_CU2|ST0_CU1);
	childregs->cp0_status &= ~(ST0_CU2|ST0_CU1);

#ifdef CONFIG_MIPS_MT_SMTC
	/*
	 * SMTC restores TCStatus after Status, and the CU bits
	 * are aliased there.
	 */
	childregs->cp0_tcstatus &= ~(ST0_CU2|ST0_CU1);
#endif
	clear_tsk_thread_flag(p, TIF_USEDFPU);

#ifdef CONFIG_MIPS_MT_FPAFF
	clear_tsk_thread_flag(p, TIF_FPUBOUND);
#endif /* CONFIG_MIPS_MT_FPAFF */

	if (clone_flags & CLONE_SETTLS)
		ti->tp_value = regs->regs[7];

	return 0;
}

/*
 * Lock a mutex (possibly interruptible), slowpath:
 */
static inline int __sched
__mutex_lock_common(struct mutex *lock, long state, unsigned int subclass,
		    struct lockdep_map *nest_lock, unsigned long ip)
{
	struct task_struct *task = current;
	struct mutex_waiter waiter;
	unsigned long flags;

	preempt_disable();
	mutex_acquire_nest(&lock->dep_map, subclass, 0, nest_lock, ip);

#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
	/*
	 * Optimistic spinning.
	 *
	 * We try to spin for acquisition when we find that there are no
	 * pending waiters and the lock owner is currently running on a
	 * (different) CPU.
	 *
	 * The rationale is that if the lock owner is running, it is likely to
	 * release the lock soon.
	 *
	 * Since this needs the lock owner, and this mutex implementation
	 * doesn't track the owner atomically in the lock field, we need to
	 * track it non-atomically.
	 *
	 * We can't do this for DEBUG_MUTEXES because that relies on wait_lock
	 * to serialize everything.
	 */

	for (;;) {
		struct task_struct *owner;

		/*
		 * If there's an owner, wait for it to either
		 * release the lock or go to sleep.
		 */
		owner = ACCESS_ONCE(lock->owner);
		if (owner && !mutex_spin_on_owner(lock, owner))
			b