Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Barry Song <21cnbao@gmail.com>
To:     gregkh@linuxfoundation.org
Cc:     linux-kernel@vger.kernel.org, linuxarm@huawei.com,
        rafael@kernel.org, robin.murphy@arm.com,
        song.bao.hua@hisilicon.com, wangzhou1@hisilicon.com,
        will@kernel.org
Subject: Re: [PATCH] platform-msi: Add ABI to show msi_irqs of platform devices
Date:   Thu, 12 Aug 2021 16:19:30 +1200
Message-Id: <20210812041930.28931-1-21cnbao@gmail.com>
In-Reply-To: <YRO5R0/N9KITjyY9@kroah.com>
References: <YRO5R0/N9KITjyY9@kroah.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

>
> > +static ssize_t platform_msi_show(struct device *dev, struct device_attribute *attr,
> > +     char *buf)
> > +{
> > + struct msi_desc *entry;
> > + unsigned long irq;
> > + int retval;
> > +
> > + retval = kstrtoul(attr->attr.name, 10, &irq);
> > + if (retval)
> > + return retval;
> > +
> > + entry = irq_get_msi_desc(irq);
> > + if (entry)
> > + return sprintf(buf, "msi\n");
>
> sysfs_emit()?

yep.

>
> But why isn't this all handled by the MSI core code?  Why would each bus
> need to have this logic in it?

i think i can extract some common code for sysfs populate/destroy to msi core from pci and platform.
but we still need some pci/platform specific code in pci-msi and platform-msi cores. for example,
pci-msi has specific data which will be accessed in its show() entry.

struct msi_desc {
        ...
        union {
                /* PCI MSI/X specific data */
                struct {
                        u32 masked;
                        struct {
                                u8      is_msix         : 1;
                                u8      multiple        : 3;
                                u8      multi_cap       : 3;
                                u8      maskbit         : 1;
                                u8      is_64           : 1;
                                u8      is_virtual      : 1;
                                u16     entry_nr;
                                unsigned default_irq;
                        } msi_attrib;
                        union {
                                u8      mask_pos;
                                void __iomem *mask_base;
                        };
                };

                ...
                struct platform_msi_desc platform;
                ...
        };
};

in addition, they are quite different in initialization/release and also need different places to save sysfs groups.
so probably i can let msi cores provide msi_populate_sysfs() and msi_destroy_sysfs() APIs. And ask pci and platform
to call msi_populate_sysfs() in their init code and save the groups in their specific pointers,  and then they can
free sysfs in their release paths by calling msi_destroy_sysfs()

>
> thanks,
>
> greg k-h

Thanks
Barry