On Thu, Mar 20 2025 at 09:23, Thomas Gleixner wrote:
> On Wed, Mar 19 2025 at 21:58, Wen Xiong wrote:
>> We don't see the issue without dynamically remove/add operation.
>> There is a small window which irqbalance daemon kicks in during device
>> reset. So it took about over 6 hours to recreate the issue when doing
>> remove/add loop operation.
>
> Sure. You need a loop to hit the window. And it does not matter whether
> it's the probe or the remove which triggers it. Fact is that the reset
> wipes out the config space, which means that any read from the config
> space between reset and restore will return garbage. That problem is not
> restricted to the interrupt code. It's a general problem.
After looking at the code again, it's a problem in the remove()
function:
__ipr_remove()
ipr_initiate_ioa_bringdown()
// resets device
restore_config_space()
....
ipr_free_all_resources()
free_irqs()
So yes, it's not probe(). But the question is pretty much the same.
Why is a reset issued while the driver is fully operational and
resources are still in use?
Don't even think about telling me that this is a problem of the MSI
interrupt rework. It is not. It's been broken forever.
You _cannot_ pull the rung under a fully operational driver and expect
that all involved parts will just magically handle this gracefully.
What about tearing down resources first and then issuing the reset?
Thanks,
tglx
