[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87fsfggone.wl-maz@kernel.org>
Date: Sat, 22 Oct 2022 11:04:37 +0100
From: Marc Zyngier <maz@...nel.org>
To: firas ashkar <firas.ashkar@...oirfairelinux.com>
Cc: alex@...riz.org.uk, tglx@...utronix.de,
linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH v2] irqchip: add TS7800v1 fpga based controller driver
On Fri, 21 Oct 2022 19:14:48 +0100,
firas ashkar <firas.ashkar@...oirfairelinux.com> wrote:
> > > > > +static void ts7800_ic_chained_handle_irq(struct irq_desc
> > > > > *desc)
> > > > > +{
> > > > > + struct ts7800_irq_data *data =
> > > > > irq_desc_get_handler_data(desc);
> > > > > + struct irq_chip *chip = irq_desc_get_chip(desc);
> > > > > + u32 mask_bits = readl(data->base + IRQ_MASK_REG);
> > > > > + u32 status = readl(data->bridge);
> > > > > +
> > > > > + chained_irq_enter(chip, desc);
> > > > > +
> > > > > + if (unlikely(status == 0)) {
> > > > > + handle_bad_irq(desc);
> > > > > + goto out;
> > > > > + }
> > > > > +
> > > > > + do {
> > > > > + unsigned int bit = __ffs(status);
> > > > > + int irq = irq_find_mapping(data->domain, bit);
> > > > > +
> > > > > + if (irq && (mask_bits & BIT(bit)))
> > > > > + generic_handle_irq(irq);
> > > >
> > > > Again, this is not appropriate. I've pointed you to the right API
> > > > in
> > > > my previous review of this patch.
> > > 'generic_handle_domain_irq' causing some issues
> >
> > Which issue?
> # insmod /tmp/virt-dma.ko
> # insmod /tmp/ts-dmac.ko
> generic_handle_domain_irq failed, error -22
> irq 2, desc: a285c39f, depth: 0, count: 0, unhandled: 0
> ->handle_irq(): e7046ee9, ts7800_ic_chained_handle_irq+0x0/0x1ec
> [irq_ts7800]
> ->irq_data.chip(): 70d81143, 0xc14db44c
> ->action(): f799c8dd
> ->action->handler(): dcf07981, bad_chained_irq+0x0/0x64
> IRQ_LEVEL set
> IRQ_NOPROBE set
> IRQ_NOREQUEST set
> IRQ_NOTHREAD set
> unexpected IRQ trap at vector 02
> generic_handle_domain_irq failed, error -22
[...]
News flash, your code is buggy. How about you debug it? The domain
name you use definitely indicates that you "know how" to do it...
M.
--
Without deviation from the norm, progress is not possible.
Powered by blists - more mailing lists