Project Home
Project Home		 Documents
Documents		 Wiki
Wiki		 Discussion Forums
Discussions		 Project Information
Project Info
QNX Operating System/Discussion/OSTech/PCI/PCIe interrupts don't work as expected/List of Posts
Forum Topic - PCI/PCIe interrupts don't work as expected: (3 Items)
   
Update
Alexander Minor
02/02/2024 7:52 AM
post122476
PCI/PCIe interrupts don't work as expected   
Hi,

I am currently porting a device driver from QNX7.1 to QNX8.0 and am observing
weird behavior (maybe a bug?) when it comes to PCI/PCIe interrupts.

When I only use PCIe devices, the interrupt handling using InterruptAttachThread()
or InterruptAttachEvent() and InterruptWait() works like expected.

When I add a PCI device to this combination, this device will stay inside
InterruptWait() and not return anymore.

I am using the x86_64 BSP and I tried it on two DELL Optiplex 9010.

Using only a single PCI device (no PCIe) in those PC's also shows different
results. In one of them it worked as expected, in the other it also stayed inside
InterruptWait().

Flags used for InterruptAttachThread/Event: _NTO_INTR_FLAGS_NO_UNMASK | _NTO_INTR_FLAGS_END.
Flags used for InterruptWait: _NTO_INTR_WAIT_FLAGS_UNMASK | _NTO_INTR_WAIT_FLAGS_FAST.

The devices use legacy interrupts, not MSI.

I attached the output of pci-tool for the working (pcie_only) and failing (pcie_and_pci)
environment. The only thing I see is, that they share an interrupt. The devices
have the vid 15cf and use IRQ 16 & 18. There is also an Intel USB Controller
that uses IRQ 16.

Are there any known PCI/PCIe issues for specific CPU/chipsets?
Does somebody know a way I could debug/resolve this issue?
Attachment: Compressed file pcie_and_pci.zip 9.98 KB
Elad Lahav
02/02/2024 8:05 AM
post122477
Re: PCI/PCIe interrupts don't work as expected   
I am not a PCI expert, but one thing to watch out for when sharing interrupts is that the source will remain masked as 
long as at least one attached handler has it masked.
You can use the new InterruptQuery() call to see if the source is masked while your driver is waiting for it.

--Elad
Alexander Minor
02/05/2024 8:19 AM
post122478
Re: PCI/PCIe interrupts don't work as expected   
HI, I tried the InterruptQuery() function and it showed that the source was masked more than once sometimes.
However, it shows the same for PCI and PCIe devices and PCIe devices do work.

I made a mistake in saying that one of the Optilex's worked. I tested it again today and it didn't. However I found a
computer in which all devices (PCI/PCIe) seem to work correctly even when interrupts are shared. Maybe it's an
issue with the hardware (DELL Optiplex 9010, Intel(R) 7 Series/C216 Chipset) or the BSP is not fully configured.

For now this works for me. Thanks for your help.

Return


  A subsidiary of BlackBerry             All content ©2004-2013, QNX Software Systems