This is the mail archive of the
ecos-discuss@sources.redhat.com
mailing list for the eCos project.
Re: Chaining DSRs in Different Drivers
- From: Jonathan Larmour <jlarmour at redhat dot com>
- To: Douglas Bush <dbush at extremeeng dot com>
- Cc: eCos Discussion List <ecos-discuss at sources dot redhat dot com>
- Date: Fri, 01 Feb 2002 17:27:44 +0000
- Subject: Re: [ECOS] Chaining DSRs in Different Drivers
- Organization: Red Hat UK Ltd.
- References: <005f01c1aabb$a9ff9470$090110ac@TRENT>
Douglas Bush wrote:
>
> Folks,
>
> Is there a good mechanism for chaining device driver DSRs?
>
> Currently, I have a PCMCIA driver for the CL-PS6700 which receives all
> the interrupts coming through the PCMCIA interface. (In this case a
> Wavelan 802.11 card.)
>
> The CL-PS6700 DSR calls the (I believe) appropriate handler, then
> re-enables interrupts. (The code is abbreviated below.)
>
> // This DSR handles the card interrupt
> static void cf_irq_dsr(cyg_vector_t vector, cyg_ucount32 count,
> cyg_addrword_t data)
> {
> struct cf_slot *slot = (struct cf_slot *)data;
>
> // Process interrupt
> (slot->irq_handler.handler)(vector, count, slot->irq_handler.param);
>
> /*
> clear all interrupts that occured in the controller
> except for CD1 and CD2
> */
> *(unsigned volatile*)PCICR = ~EDB7XXX_CF_DETECT_MASK;
>
> // Clear interrupt
> cyg_drv_interrupt_acknowledge(EDB7XXX_CF_IRQ);
>
> // Allow interrupts to happen again
> cyg_drv_interrupt_unmask(EDB7XXX_CF_IRQ);
> }
>
> The (slot->irq_handler.handler)(xxx) call actually calls eth_drv_dsr
> (bottom of the eth_drv.c file) which schedules a timer but doesn't
> actually call the Wavelan driver DSR until sometime later.
>
> The PCMCIA DSR then re-enables interrupts to the PCMCIA port BEFORE the
> Wavelan DSR has run, the PCMCIA driver therefore goes into near
> continuous interrupt.
>
> Is there a better mechanism for chaining DSRs from one driver (PCMCIA)
> to another (Ethernet)? Preferably from the context of the original DSR?
Well as you can see, not at present :-). Perhaps the responsibility for
clearing and re-enabling interrupts should be passed down to the next level
(i.e. slot->irq_handler.handler, which could then also be passed the IRQ to
clear..... or perhaps instead passed a little callback function so that the
lower layers can call back into the cf level code to clear the interrupt.
I wonder what Gary thinks?
Jifl
--
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine