Messages in this thread | | | Date | Wed, 10 Sep 2008 21:44:09 +0400 | From | Cyrill Gorcunov <> | Subject | Re: [PATCH] x86: io-apic - get rid of __DO_ACTION macro |
| |
[Yinghai Lu - Wed, Sep 10, 2008 at 10:30:39AM -0700] | On Wed, Sep 10, 2008 at 2:31 AM, Ingo Molnar <mingo@elte.hu> wrote: | > | > * Yinghai Lu <yhlu.kernel@gmail.com> wrote: | > | >> On Tue, Sep 9, 2008 at 11:02 PM, Cyrill Gorcunov <gorcunov@gmail.com> wrote: | >> > On Wed, Sep 10, 2008 at 12:13 AM, Yinghai Lu <yhlu.kernel@gmail.com> wrote: | >> > ... | >> >> | >> >> hope we can keep using MACRO.. | >> >> | >> >> YH | >> >> | >> > | >> > Btw, Yinghai, what does it mean? To not touch this macro at all? | >> > Or you mean about implementation issue (ie the design itself)? | >> | >> do not touch this macro... and may revisit after 2.6.28 | > | > anything you are particularly worried about? Regressions we should be | > able to find pretty quickly, in a central macro like that - and the | > macro is quite ugly. | > | | ok, let remove unneeded "if", and use function pointer... | | void (*extra_action_t)(struct irq_pin_list *entry); | | +static inline void io_apic_modify_irq(unsigned int irq, | + int mask_and, int mask_or, | + int mask_and_not, extra_action_t action) | +{ | + int pin; | + struct irq_cfg *cfg; | + struct irq_pin_list *entry; | + cfg = irq_cfg(irq); | + for (entry = cfg->irq_2_pin; entry != NULL; entry = entry->next) { | + unsigned int reg; | + pin = entry->pin; | + reg = io_apic_read(entry->apic, 0x10 + pin * 2); | + reg &= mask_and; | + reg |= mask_or; | + reg &= ~mask_and_not; | + io_apic_modify(entry->apic, 0x10 + pin * 2, reg); | + if (action) | + action(entry); | + } | +} | | +void extra_read(struct irq_pin_list *entry) | + { | + /* | + * Synchronize the IO-APIC and the CPU by doing | + * a dummy read from the IO-APIC | + */ | + struct io_apic __iomem *io_apic; | + io_apic = io_apic_base(entry->apic); | + readl(&io_apic->data); | + } | | | YH |
Yinghai, I've posted second version before you proposed this. Let me update it then. (though I don't think if we would need some extra actions instead of syncs by additional read except for possible erranious chips). Anyway - will repost updated version. Thanks.
- Cyrill -
| |