Re: [PATCH 3/4] PCI: endpoint: pci-epf-test: Don't free doorbell IRQ unless requested

On Mon, Feb 16, 2026 at 11:30:44PM +0900, Koichiro Den wrote:
> On Mon, Feb 16, 2026 at 12:35:56PM +0100, Niklas Cassel wrote:
> > On Mon, Feb 16, 2026 at 12:09:13AM +0900, Koichiro Den wrote:
> > > pci_epf_test_enable_doorbell() allocates a doorbell and then installs
> > > the interrupt handler with request_threaded_irq(). On failures before
> > > the IRQ is successfully requested (e.g. no free BAR,
> > > request_threaded_irq() failure), the error path jumps to
> > > err_doorbell_cleanup and calls pci_epf_test_doorbell_cleanup().
> > > 
> > > pci_epf_test_doorbell_cleanup() unconditionally calls free_irq() for the
> > > doorbell virq, which can trigger "Trying to free already-free IRQ"
> > > warnings when the IRQ was never requested or when request_threaded_irq()
> > > failed.
> > > 
> > > Track whether the doorbell IRQ has been successfully requested and only
> > > call free_irq() when it has.
> > > 
> > > Fixes: eff0c286aa91 ("PCI: endpoint: pci-epf-test: Add doorbell test support")
> > > Signed-off-by: Koichiro Den <[email protected]>
> > > ---
> > >  drivers/pci/endpoint/functions/pci-epf-test.c | 9 ++++++++-
> > >  1 file changed, 8 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/pci/endpoint/functions/pci-epf-test.c b/drivers/pci/endpoint/functions/pci-epf-test.c
> > > index 148a34e51f6b..defe1e2ea427 100644
> > > --- a/drivers/pci/endpoint/functions/pci-epf-test.c
> > > +++ b/drivers/pci/endpoint/functions/pci-epf-test.c
> > > @@ -86,6 +86,7 @@ struct pci_epf_test {
> > >  	bool			dma_private;
> > >  	const struct pci_epc_features *epc_features;
> > >  	struct pci_epf_bar	db_bar;
> > > +	bool			db_irq_requested;
> > 
> > It would be nice if we could avoid this, it looks a bit odd.
> > 
> > 
> > >  	size_t			bar_size[PCI_STD_NUM_BARS];
> > >  };
> > >  
> > > @@ -715,7 +716,10 @@ static void pci_epf_test_doorbell_cleanup(struct pci_epf_test *epf_test)
> > >  	struct pci_epf_test_reg *reg = epf_test->reg[epf_test->test_reg_bar];
> > >  	struct pci_epf *epf = epf_test->epf;
> > >  
> > > -	free_irq(epf->db_msg[0].virq, epf_test);
> > > +	if (epf_test->db_irq_requested && epf->db_msg) {
> > > +		free_irq(epf->db_msg[0].virq, epf_test);
> > > +		epf_test->db_irq_requested = false;
> > > +	}
> > >  	reg->doorbell_bar = cpu_to_le32(NO_BAR);
> > >  
> > >  	pci_epf_free_doorbell(epf);
> > > @@ -732,6 +736,8 @@ static void pci_epf_test_enable_doorbell(struct pci_epf_test *epf_test,
> > >  	size_t offset;
> > >  	int ret;
> > >  
> > > +	epf_test->db_irq_requested = false;
> > > +
> > >  	ret = pci_epf_alloc_doorbell(epf, 1);
> > >  	if (ret)
> > >  		goto set_status_err;
> > > @@ -751,6 +757,7 @@ static void pci_epf_test_enable_doorbell(struct pci_epf_test *epf_test,
> > >  		goto err_doorbell_cleanup;
> > >  	}
> > >  
> > > +	epf_test->db_irq_requested = true;
> > >  	reg->doorbell_data = cpu_to_le32(msg->data);
> > >  	reg->doorbell_bar = cpu_to_le32(bar);
> > >  
> > 
> > Can't we do something like:
> > 
> > -For all goto's after request_threaded_irq() success case:
> > jump to a label that also cleans up the IRQ.
> > 
> > For failures before or at request_threaded_irq(), jump to
> > a label that does not call free_irq().
> 
> I thought this would be a minimal change to avoid the problematic case, but I
> agree it's not very clean (the "db_irq_requested" flag looks a bit odd).
> 
> So I'll split pci_epf_test_doorbell_cleanup() into two helper functions to match
> your suggested structure.
> 
> > 
> > 
> > 
> > pci_epf_test_disable_doorbell() should probably return error
> > if (!epf_test->db_bar.size)
> > 
> > (before pci_epf_test_disable_doorbell() calls free_irq())
> > 
> > pci_epf_test_disable_doorbell() should probably also memset
> > epf_test->db_bar.
> 
> and I'll take care of these suggestions as well.

I revisited the latest doorbell code with those latter suggestions in mind.
Looking at the current flow, it seems we may not actually need these additional
safeguards.

- any error path inside pci_epf_test_enable_doorbell() resets reg->doorbell_bar
  to NO_BAR.
- pci_epf_test_disable_doorbell() performs its cleanup only if reg->doorbell_bar
  is not NO_BAR. When it runs, it always resets reg->doorbell_bar to NO_BAR, so
  it cannot effectively run twice.

Given that the guarding logic is already centralized around reg->doorbell_bar,
explicitly clearing epf_test->db_bar does not seem strictly necessary.

It would be more of a refactoring cleanup than a functional fix.

Koichiro

> 
> Thanks for the review,
> Koichiro
> 
> > 
> > 
> > Kind regards,
> > Niklas