On Tue, Feb 17, 2026 at 11:15:08AM +0100, Niklas Cassel wrote:

Hello Koichiro,

On Tue, Feb 17, 2026 at 05:06:00PM +0900, Koichiro Den wrote:

quoted

pci-epf-test advertises the doorbell target to the RC as a BAR number
and an offset, and the RC rings the doorbell with a single DWORD MMIO
write.

Some doorbell backends may report that the doorbell target is already
exposed via a platform-owned fixed BAR (db_msg[0].bar/offset). In that
case, reuse the pre-exposed window and do not reprogram the BAR with
pci_epc_set_bar().

Also honor db_msg[0].irq_flags when requesting the doorbell IRQ, and
only restore the original BAR mapping on disable if pci-epf-test
programmed it.

Reviewed-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Koichiro Den <redacted>
---

(snip)

quoted

@@ -753,22 +771,33 @@ static void pci_epf_test_enable_doorbell(struct pci_epf_test *epf_test,
 	reg->doorbell_data = cpu_to_le32(msg->data);
 	reg->doorbell_bar = cpu_to_le32(bar);
 
-	msg = &epf->db_msg[0].msg;
-	ret = pci_epf_align_inbound_addr(epf, bar, ((u64)msg->address_hi << 32) | msg->address_lo,
-					 &epf_test->db_bar.phys_addr, &offset);
+	if (db->bar == NO_BAR) {
+		ret = pci_epf_align_inbound_addr(epf, bar,
+						 ((u64)msg->address_hi << 32) |
+						 msg->address_lo,
+						 &epf_test->db_bar.phys_addr,
+						 &offset);
 
-	if (ret)
-		goto err_free_irq;
+		if (ret)
+			goto err_free_irq;
+	}

I tried this series on Rock5b (RK3588), and was surprised to see the doorbell
test case still failing.

Thank you very much for testing, and apologies for not being able to test on
RK3588 on my side right now.

quoted

+
+	if (size_add(offset, sizeof(u32)) > epf->bar[bar].size)
+		goto err_doorbell_cleanup;

It turns out that this check is the reason for it still failing.

You see, for a BAR that is marked as BAR_RESERVED, pci-epf-test will not
allocate backing memory, so epf->bar[bar].size will be 0.

If I removed this check, I could get the test case to pass.

As I suggested in my previous email, perhaps this check is better suited
in pci_epf_alloc_doorbell(). (As a DWORD alignment check inside
pci_epf_alloc_doorbell(). pci_epf_alloc_doorbell() could itself return
error if the doorbell is not DWORD aligned.)

Yes, and I mentioned there I would reconsider this when respinning this feature
series. For reference:
https://lore.kernel.org/linux-pci/jcjson6zedvhkpctwzfao2wfaaujtdfqsnnm3k25e2vpz2evf4@hbsegnyevisu/ (local)
I hadn't noticed the bar.size==0 possibility at that time.

I agree that this check is better placed in pci_epf_alloc_doorbell().
I'll respin accordingly.

Thanks again for the testing and review,
Koichiro

That way, you could remove this check from pci_epf_test_enable_doorbell(),
and we don't need to care about epf->bar[bar].size.


Kind regards,
Niklas