Message-ID: <87ecwdqct9.ffs@tglx>
Date: Sun, 25 May 2025 11:06:42 +0200
From: Thomas Gleixner <tglx@...utronix.de>
To: Tianyang Zhang <zhangtianyang@...ngson.cn>, chenhuacai@...nel.org,
 kernel@...0n.name, corbet@....net, alexs@...nel.org, si.yanteng@...ux.dev,
 jiaxun.yang@...goat.com, peterz@...radead.org, wangliupu@...ngson.cn,
 lvjianmin@...ngson.cn, maobibo@...ngson.cn, siyanteng@...oftware.com.cn,
 gaosong@...ngson.cn, yangtiezhu@...ngson.cn
Cc: loongarch@...ts.linux.dev, linux-doc@...r.kernel.org,
 linux-kernel@...r.kernel.org, Tianyang Zhang <zhangtianyang@...ngson.cn>
Subject: Re: [PATCH v3 2/2] irq/irq-loongarch-ir:Add Redirect irqchip support

On Fri, May 23 2025 at 18:18, Tianyang Zhang wrote:
>  
> -static void avecintc_sync(struct avecintc_data *adata)
> +void avecintc_sync(struct avecintc_data *adata)
>  {
>  	struct pending_list *plist;
>  
> @@ -109,7 +99,7 @@ static int avecintc_set_affinity(struct irq_data *data, const struct cpumask *de
>  			return -EBUSY;
>  
>  		if (cpu_online(adata->cpu) && cpumask_test_cpu(adata->cpu, dest))
> -			return 0;
> +			return IRQ_SET_MASK_OK_DONE;

This change really wants to be seperate with a proper explanation and
not burried inside of this pile of changes.

> +static inline bool invalid_queue_is_full(int node, u32 *tail)
> +{
> +	u32 head;
> +
> +	head = read_queue_head(node);

Please move the initialization into the declaration line:

       u32 head = read_queue...();

All over the place, where it's the first operation in the code. That
makes the code more dense and easier to follow.

> +	*tail = read_queue_tail(node);
> +
> +	return !!(head == ((*tail + 1) % INVALID_QUEUE_SIZE));

What's the !! for? A == B is a boolean expression already.

> +}
> +
> +static void invalid_enqueue(struct redirect_queue *rqueue, struct irde_inv_cmd *cmd)
> +{
> +	struct irde_inv_cmd *inv_addr;
> +	u32 tail;
> +
> +	guard(raw_spinlock_irqsave)(&rqueue->lock);
> +
> +	while (invalid_queue_is_full(rqueue->node, &tail))
> +		cpu_relax();
> +
> +	inv_addr = (struct irde_inv_cmd *)(rqueue->base + tail * sizeof(struct irde_inv_cmd));
> +	memcpy(inv_addr, cmd, sizeof(struct irde_inv_cmd));
> +	tail = (tail + 1) % INVALID_QUEUE_SIZE;
> +
> +	/*
> +	 * The uncache-memory access may have an out of order problem cache-memory access,
> +	 * so a barrier is needed to ensure tail is valid
> +	 */

This comment does not make sense at all.

What's the actual uncached vs. cached access problem here? AFAICT it's
all about the ordering of the writes:

    You need to ensure that the memcpy() data is visible _before_ the
    tail is updated, no?

> +	wmb();
> +
> +	write_queue_tail(rqueue->node, tail);
> +}

> +static int redirect_table_free(struct redirect_item *item)

That return value is there to be ignored by the only caller, right?

> +{
> +	struct redirect_table *ird_table;
> +	struct redirect_entry *entry;
> +
> +	ird_table = item->table;
> +
> +	entry = item->entry;
> +	memset(entry, 0, sizeof(struct redirect_entry));
> +
> +	scoped_guard(raw_spinlock_irqsave, &ird_table->lock)
> +		bitmap_release_region(ird_table->bitmap, item->index, 0);
> +
> +	kfree(item->gpid);
> +
> +	irde_invlid_entry_node(item);
> +
> +	return 0;
> +}
> +
> +static inline void redirect_domain_prepare_entry(struct redirect_item *item,
> +					struct avecintc_data *adata)

Please align the argument in the second line properly:

https://www.kernel.org/doc/html/latest/process/maintainer-tip.html#line-breaks

> +
> +static inline void redirect_ack_irq(struct irq_data *d)
> +{
> +}
> +
> +static inline void redirect_unmask_irq(struct irq_data *d)
> +{
> +}
> +
> +static inline void redirect_mask_irq(struct irq_data *d)
> +{
> +}

These want some explanation why they are empty.

> +
> +static struct irq_chip loongarch_redirect_chip = {
> +	.name			= "REDIRECT",
> +	.irq_ack		= redirect_ack_irq,
> +	.irq_mask		= redirect_mask_irq,
> +	.irq_unmask		= redirect_unmask_irq,
> +	.irq_set_affinity	= redirect_set_affinity,
> +	.irq_compose_msi_msg	= redirect_compose_msi_msg,
> +};
> +out_free_resources:
> +	redirect_free_resources(domain, virq, nr_irqs);
> +	irq_domain_free_irqs_common(domain, virq, nr_irqs);
> +
> +	return -EINVAL;

-ENOMEM?

> +}
> +
> +	bitmap = bitmap_zalloc(IRD_ENTRIES, GFP_KERNEL);
> +	if (!bitmap) {
> +		pr_err("Node [%d] redirect table bitmap alloc pages failed!\n", node);
> +		return -ENOMEM;

Leaks pages.

> +	}
> +
> +	ird_table->bitmap = bitmap;
> +	ird_table->nr_ird = IRD_ENTRIES;
> +	ird_table->node = node;
> +
> +	raw_spin_lock_init(&ird_table->lock);
> +
> +	if (redirect_queue_init(node))
> +		return -EINVAL;

Leaks pages and bitmap.

> +
> +	iocsr_write64(CFG_DISABLE_IDLE, LOONGARCH_IOCSR_REDIRECT_CFG);
> +	iocsr_write64(__pa(ird_table->table), LOONGARCH_IOCSR_REDIRECT_TBR);
> +
> +	return 0;
> +}

> +#if defined(CONFIG_ACPI)

#ifdef CONFIG_ACPI

> +static int __init redirect_reg_base_init(void)
> +{
> +	acpi_status status;
> +	uint64_t addr = 0;

What's this initialization for?

> +int __init redirect_acpi_init(struct irq_domain *parent)
> +{
> +	struct fwnode_handle *fwnode;
> +	struct irq_domain *domain;
> +	int ret;
> +
> +	fwnode = irq_domain_alloc_named_fwnode("redirect");
> +	if (!fwnode) {
> +		pr_err("Unable to alloc redirect domain handle\n");
> +		goto fail;
> +	}
> +
> +	domain = irq_domain_create_hierarchy(parent, 0, IRD_ENTRIES, fwnode,
> +			&redirect_domain_ops, irde_descs);

Please align the arguments in the second line properly.

> +	if (!domain) {
> +		pr_err("Unable to alloc redirect domain\n");
> +		goto out_free_fwnode;
> +	}
> +
> +	redirect_domain = domain;
> +
> +	ret = redirect_table_init(0);
> +	if (ret)
> +		goto out_free_table;
> +
> +	ret = acpi_cascade_irqdomain_init();
> +	if (ret < 0) {
> +		pr_err("Failed to cascade IRQ domain, ret=%d\n", ret);
> +		goto out_free_table;
> +	}
> +
> +	cpuhp_setup_state_nocalls(CPUHP_AP_IRQ_REDIRECT_STARTING,
> +				  "irqchip/loongarch/redirect:starting",
> +				  redirect_cpu_online, NULL);

Hmm.

> +static int redirect_cpu_online(unsigned int cpu)
> +{
> +	int ret, node = cpu_to_node(cpu);
> +
> +	if (cpu != cpumask_first(cpumask_of_node(node)))
> +		return 0;
> +
> +	ret = redirect_table_init(node);
> +	if (ret) {
> +		redirect_table_fini(node);
> +		return -EINVAL;
> +	}
> +
> +	return 0;
> +}

So if you unplug all CPUs of a node and then replug the first CPU in the
node, then this invokes redirect_table_init() unconditionally, which
will unconditionally allocate pages and bitmap again ....

Thanks,

        tglx

Powered by blists - more mailing lists